News Feeds

A recent study from Open Plans and the Transportation Research Program at Hunter College argued that AVs will generate excessive “deadheading,” or empty vehicle miles, in New York City. The concern is fair, but the conclusion was based on a false equivalence: it treats the density of California as directly comparable to New York City. The literature is clear that deadheading depends on demand density. There are fewer empty miles when cars don’t have to travel far to find their next passenger.

Nonetheless, the report raises an important question: how can AVs avoid creating more congestion? It’s a fair question, given the rapid introduction of tens of thousands of personally owned for-hire vehicles more than a decade ago was a shock to the system. While AVs will have a much smaller footprint, there are lessons we all can learn from mistakes made by companies such as Uber and Lyft.

I walk and bike everywhere and I want the same for my kids. That’s why I was thrilled that after a year of advocacy and organizing, the City of Oakland is redesigning one of the city’s most dangerous streets to prioritize bicycle and pedestrian safety. It’s also why I work at Waymo. In my personal advocacy and in my role as the Policy Development and Research Manager at Waymo, I often think about the trade-offs required to build a safe, vibrant, and efficient transportation system. Whether it is bike or bus lanes, curb management, or parking, there is rarely a perfect solution. The conversation about how to most effectively deploy AVs is no different, and curb management is an important place to start.

Cities become more delightful when there are fewer cars circling and less curb space devoted to vehicle storage. Right now, roughly 40 percent of ride-hailing miles are empty, a figure that generally holds for AV fleets. But as a fleet operator, Waymo is incentivized to minimize deadheading because we bear the cost of every single empty mile driven. Every unnecessary mile means additional operating costs, energy consumption, vehicle wear, and risk exposure. Traditional ride-hailing platforms don’t internalize these costs; they are pushed onto the drivers. Waymo, like cities, wants to reduce empty miles as much as possible.

One solution to reducing empty miles is parking between rides. Taxis do this at taxi stands, and the San Francisco County Transportation Authority found ride-hail drivers park an average of 10 minutes between hails. When Waymo asked researchers from UC Berkeley and UC Irvine to model the impact on VMT from AVs staging at the curb, they found that it reduced empty miles by more than 60 percent. And the magnitude of the reduction was striking, particularly because the model strategically accounted for moving empty AVs into off-street lots during low-demand periods.

Staging is also a highly productive use of curb space, according to the researchers. Because AVs spend much of their time providing trips, the cumulative value generated by staging far exceeds that of a curb space occupied by a privately owned vehicle. For a commercial district, this means that local businesses could receive eight to 10 times as many potential customer visits than if the space were designated for private vehicle storage only.

None of this means AVs should have unrestricted access to the curb. Waymo already prioritizes parking in off-street lots, limits itself to one staged vehicle per block face, and avoids cycling through the same spot repeatedly. In a recent report, Open Plans and the Transportation Research Program at Hunter College argued that New York City’s “constrained curb space” makes staging infeasible, leaving AVs with little choice but to drive empty between trips. But the curb isn’t constrained by the laws of physics, it’s constrained by policy. As the Berkeley and Irvine researchers suggested, establishing permitting standards can formalize curbside staging. Deadheading can never be zero, so the real question is how many empty miles will we accept in exchange for the safety, accessibility and other benefits of AVs. Curb management is exactly how cities can organize that tradeoff with clear rules.

Encouragingly, New York is having these conversations at the highest level. Mayor Zohran Mamdani created an Office of Curb Management this year to bring more order to how the city allocates curb space. Officials are weighing ideas from metered spots to residential permits to dynamic pricing. Each plan differs, but the premise is one we agree with: curb space is valuable, and managing it inconsistently doesn’t serve anyone well.

Waymo, city officials, and urbanists have common cause, and we don’t have to wait to start having these conversations. Open Plans has in fact been a leader on reimagining curbs. New York is the first and only city in the nation to implement congestion pricing, which has been a resounding success. Though controversial, there is renewed interest in more visibly holding drivers accountable for blocking street cleaning.

At Waymo, we believe AVs can play a valuable role in making transportation safer and cities more livable. Conversations about curb management and other policies that help responsibly deploy AVs are key to our work with the city officials, advocates, and communities we hope to serve.

• Congressional Democrats are accusing the Federal Transit Administration of intentionally stonewalling transit projects. The Trump administration has yet to issue any grants from a multibillion-dollar capital investment program since the president’s second term started. It’s also tried to cut off funding for already-approved projects in blue states, like the Gateway Tunnel. (New York. Times)
• Transportation Secretary Sean Duffy believes bike lanes cause traffic congestion. In fact, the opposite is true. (Bloomberg Government)
• Despite tariffs, demand for new transit buses in North America remains strong as agencies replace their aging fleets. (Smart Cities Dive)
• Safety and comfort are just as important as distance when it comes to walkability. (The Conversation)
• The privately owned cross-country passenger rail line AmeriStarRail has yet to materialize because the company hasn’t been able to reach an agreement with Amtrak. (USA Today)
• Los Angeles Mayor Karen Bass’ campaign promise to make the L.A. Metro free would cost $1 billion and possibly prevent the agency from improving service. (LAist)
• A Washington, D.C. cycling group is pushing for better walking, biking and transit access around the new RFK stadium. (Greater Greater Washington)
• Portland residents who bought an e-bike using the city’s rebate program say the purchase changed their lives. (Mercury)
• Kansas City is using water-spraying trucks to keep extreme heat from warping streetcar tracks. (Axios)
• The Metro Atlanta Rapid Transit Authority has three finalists for CEO. (Saporta Report)
• A Vision Zero plan in the Seattle suburb of Lynnwood is getting pushback from people who say the city doesn’t need to spend money on traffic safety because it already has relatively few road deaths. (The Urbanist)
• Ann Arbor transit agency The Ride is starting a shuttle service for University of Michigan football games. (MLive)
• Heavy electric vehicles are tearing up China’s roads. (Techy 44)
• A proposal for high-speed rail in Australia would save a lot of jet fuel and help solve Sydney’s affordable housing crisis. (B1M)
• Japanese cities are so livable because everyday needs are within walking distance, and there’s an excellent transit system for longer trips. (BBC)

ALERT’s citizen suit clears the way for safer products during oil spill response

August 13, 2026 – Recently, an Oregon state oil spill response planner asked me, “Have you seen EPA’s new product schedule?” I had not. “Something weird happened to it,” he said. “There’s almost nothing on it.” We both googled it and discovered that EPA’s new list of products for oil spill response contained five products. “See?” the state planner said. “What happened to the other 130 products?”

What happened was the successful lawsuit brought by ALERT and allies in 2020. Our citizen suit changed the rules governing use of dispersants (our focus) and other products – such as surface washing agents, solidifiers, herding agents, and bioremediation agents – during oil spill responses. EPA tightened the protocols for testing products to eliminate more toxic products, required full disclosure of ingredients (no hiding toxic ingredients as “confidential business information”), and significantly, added rules for removing products from the schedule –  an option previously unavailable to citizens. Rules for removal require evidence that technical information supplied to EPA by the manufacturer was incorrect, inaccurate, misleading, or outdated concerning impacts to human health and the environment. In other words, anything less than truthful.

Under court supervision, EPA allowed a lengthy three-year conditional use period, counting the six-months between promulgation of the new rules and when they went into effect on December 11, 2023. [88 FR 38336, June 12, 2023] That conditional use period – which grandfathered in all 130+ products that were listed as of June 12, 2023 – ended on June 10, 2026. The new EPA Product Schedule lists four dispersants and one surface washing agent. Apparently, the manufacturers of the other 130+ products could not or chose not to meet the new requirements.

Was it all smoke and mirrors? Decades of deception that oil spills can be “cleaned up” with chemical products that miraculously create a net benefit instead of more harm than good? Manipulated lab tests? Biased statistics and formulas? Carefully worded labeling that could make black appear to be white? A growing “wealth defense industry” with legions of lawyers who could make thousands of toxic tort cases disappear from workers and the public who were directly exposed to their product? Apparently so.

The famous “Gulf Walrus” are another myth found only in oil spill contingency plans for the Gulf of Mexico at the time of the Deepwater Horizon oil disaster. The error was corrected but the industry’s disrespect for human health and the environment remains.

Case in point: Corexit dispersants were the go-to in the US for oil spill response and comprised 45% of global stockpiles. No more. The former owner of Corexit dispersants, ChampionX, voluntarily pulled the plug on its Corexit product line in November 2022 when EPA’s proposed rules were moving inexorably under court supervision towards promulgation. Then, ChampionX vanished altogether when it was acquired by SLB (Schlumberger) in 2025. Corexit was just one of the 130+ products. The others just disappeared more quietly on June 10, 2026.

We did this – ALERT and the hundreds of concerned citizens over fifteen years who stepped up and signed onto petitions for this rulemaking and the lawsuit. We wiped out 130+ likely toxic products on EPA’s list.

So, where does this leave us?ALERT and the Government Accountability Project filed a noncompliance complaint with OSHA about its less than truthful reporting on technical literature for three of the five dispersants that are currently on EPA’s list. (The other two products were listed after we filed our complaint.) Our complaint is still under active investigation.

At this tipping point, we remain committed to eliminating products that will cause more harm than good during oil spill response. Why? Beyond the obvious, holding the oil industry accountable for multi-organ harm across body systems and species including humans from exposure to its products helps build pressure towards leaving the oil in the ground and transitioning to independence from fossil fuels.

In solidarity,

 by Adm. Chris Barrie (Rtd), first published at The Diplomat 

We live in a rapidly evolving risk landscape. The conflict in the Middle East has not resolved and its effects will last for months, if not years. The Iran-U.S. clashes have transformed into something that strains supply chains, reduces crop yields, rewrites regional politics, and concentrates attention on near-term stability at the expense of the big global challenges.

The decisions by the Trump administration to retreat from the commitments and frameworks of global climate action have created a leadership void that is felt most acutely not in Washington or Brussels, but in the low-lying islands of the Pacific, in the Australian communities already contending with extreme fire seasons and more intense floods and cyclones, and in every boardroom and cabinet room where 30-year infrastructure decisions are being made in the face of deepening uncertainty.

The decline of America as a global leader represents a startling shift in the security space Australia has occupied. The Trump administration’s attack on international conventions and relationships, and its rejection of values that many Australians hold dear, requires Australia to rethink how it exercises leadership and influence in our region.

As someone who has spent over two decades arguing – from inside defense and intelligence institutions as much as outside them – that climate disruption is the most significant strategic threat Australia faces, this is a moment for clear-eyed reckoning.

The world has become more dangerous, and climate disruption adds to the danger. The too-slow pace of the energy transition and faster-than-forecast climate impacts are escalating risks. 

For the better part of my career in uniform, climate change was treated as a secondary issue in defense planning – something that belonged to the environment portfolio, not to the strategic one.

We now have more than two decades of analyses – from defense and intelligence agencies in Australia and across allied nations, from the Australian Security Leaders Climate Group, and from the assessments of Australia’s Treasury department and Productivity Commission – all pointing to the same conclusion. The risks of climate disruption are not linear, but compounding and cascading. When one system fails – whether that is a power grid, a water catchment, an agricultural region, or a community’s capacity for self-recovery – it puts pressure on every connected system.

We must redesign critical systems to make those systems more resilient, more sovereign, and less vulnerable to the shocks that are coming.

The choices Australia makes in the next five years about the pace and shape of the energy transition will determine not only our electricity prices, but our strategic position – our credibility in the region, our leverage in international coalitions, and our capacity to lead rather than follow. That is why I no longer think of the energy transition primarily as climate mitigation policy but as nation-building.

Australia’s energy transition broke records in 2025. Renewables generated almost half our electricity. Battery storage capacity tripled. This performance is a genuine achievement.

The pipeline of projects is real, but the delivery gap between ambition and projects on the ground remains large.

Policy consistency cannot stop at renewable energy. A nation cannot simultaneously ask this industry to build the clean-energy economy of the future, while continuing to expand the industries driving climate disruption.

Renewables are half the story, but phasing out fossil fuels is the other, and from that standpoint Australian policy is contradictory, perhaps even delusional.

The Australian government has a clear policy of talking about and championing renewable energy, while avoiding, wherever possible, any discussion of climate impacts and how Australian coal and gas, including new licenses, are helping drive the world toward an unliveable planet.

Our climate minister, Chris Bowen, has said Australia will not use its role as president of the negotiations at the 2026 COP31 climate summit to facilitate an agreement to curtail coal or gas exports, declaring in a recent media interview that it is “not on the agenda” under his leadership.

The Australian Security Leaders Climate Group previously documented how the government has found confronting climate impacts and their security consequences simply “too hot to handle.”

Whatever short-term income fossil fuel exports generate is overwhelmed by the deaths, destruction, and decline in human security caused by accelerating climate disruption. These costs will continue accumulating for decades and be borne by communities, governments and future generations.

The World Health Organization’s estimate that extreme heat has claimed more than 200,000 lives in Europe since 2022, with millions more affected physically and mentally, is an example of the human consequences already unfolding.

It is a consistent government practice to avoid discussion of the science, the impacts, and Australia’s outsized contribution to them.

We see a similar pattern in the United States, where a faction of Congressional Democrats have embraced what has been called “climate hushing” – scaling back discussion of climate change and shifting the conversation toward energy affordability.

In times of crisis, especially when the future security of the nation is at stake, leadership means confronting the hardest problems head on, as Australia has done during times of war and during the COVID-19 pandemic.

Taking difficult issues off the agenda does not demonstrate leadership. It demonstrates hesitation. Sweeping problems under the carpet exhibits weakness. And outright denial looks like cowardice to me, wherever it comes from. 

We cannot credibly claim to be addressing the climate emergency while simultaneously increasing the very emissions driving it. Australia’s fossil fuel export trajectory is like trying to empty a bathtub while leaving the tap running. Every new fossil fuel project and every expansion in exports cancels out part of the global progress being delivered by the clean energy industry. 

We are not short of technology. We are increasingly not short of public support. What we are short of is institutions willing to hold the line once the smoke clears and the cameras move on.

This is a re-post from The Climate Brink

AI energy use is a huge and controversial topic at the moment. Credible estimates have AI data centers accounting for around 12% US electricity use by 2030. But at the same time consumers have been given reassuringly small numbers about the impact of their own AI use, numbers that seem on their face somewhat inconsistent with the staggering size of their aggregate usage.

In 2025 Google published an article calculating that median Gemini text prompt used only 0.24 watt-hours (Wh), less energy than “watching nine seconds of television”. Around the same time, Sam Altman said that an average ChatGPT query uses about 0.34 Wh, and Epoch AI came out with similar numbers. Writers like Andy Masley and Hannah Ritchie have shown that at these rates an individual using chatbots has a pretty negligible impact, with one prompt only amounting to roughly 1/150,000th of an average American’s daily emissions.

Those numbers are basically right. They are also increasingly divorced from how AI is actually being used today.

The fastest-growing way that software engineers and scientists actually use AI is not typing questions into a chat box. Rather, we use AI agents through tools like Claude Code and Codex that plan, write code, run it, read the results, and iterate on their own. These agents make dozens of model calls per human prompt, and engage in complex reasoning chains that involve attempting and evaluating multiple answers to the same question.

I work for a company in Silicon Valley (Stripe) and admittedly use the latest AI tools more than most people. But I thought it would be instructive to take a deep dive into my own AI use over the past 8 weeks and calculate the actual energy use I was responsible for.

Over the past 8 weeks I typed 1,138 prompts into Claude Code. Those prompts triggered more than 14,000 model calls that processed 3.2 billion tokens. My best estimate is that this used around 170 kWh of data center electricity (with an uncertainty range of roughly 70 to 330 kWh across methods and assumptions). That works out to around 150 Wh per prompt (60 to 290 Wh), which is roughly 600 times (250 to 1,200) the energy of a median chat prompt. A “prompt” is ultimately not a unit of AI use any more than “trips” is a measurement of driving; it’s how far you go that matters.

Agents supercharge AI usage

Part of the impetus for this post is the publication of a new white paper from Watershed (Bistline et al. 2026) proposing a standardized framework for corporate AI emissions accounting. It is the most careful treatment I have seen of why published per-query numbers differ by orders of magnitude (system boundaries, mostly), and it contains a figure that should reframe the whole discussion: electricity per AI task spans more than five orders of magnitude, from thousandths of a watt-hour for text classification to 50-500 Wh for an agentic workflow making 5-50 frontier model calls. As they put it, emissions attributed to one “interaction” may understate the compute actually consumed “by an order of magnitude or more.”

Other researchers have found similar results. Bai et al. (2026) measured coding agents on real software tasks and found they consume roughly 1,000 times the tokens of an ordinary chatbot interaction. And these sort of agents tasks represent the most rapid driver of increased AI usage; Anthropic’s Economic Index found that 97% of their API usage now show “automation-dominant” patterns associated with agents.

To put these values in perspective, the figure below compares published per-prompt and task estimates (blue) with what I measured from my Claude Code use (orange) as well as common benchmarks for energy use (running a microwave, a fridge, or a whole home):

Electricity consumption per AI task, including published estimates (blue) and values computed from my own Claude Code session logs (orange). Measured token counts converted using Bistline (2026) activity-tier energy factors; orange ranges span cache-read energy assumptions of 1% to 25%.

My median Claude Code session uses around 0.6 kWh (0.25 to 1.2 kWh), which is at the top end of Watershed’s generic agentic usage estimate, and fifty times the energy used to charge a cellphone. My average day of Claude Code (3.0 kWh, range 1.2 to 5.9 kWh) uses more electricity than running two refrigerators.

Measuring my own footprint

Claude Code keeps complete local transcripts of every session, including the exact token counts the API reports for every model call.1 This lets me precisely know how much AI usage I was responsible for rather than simply extrapolating it from published benchmarks; its only the step to convert tokens used to energy that requires assumptions.

The first thing I found is that the gap between “prompts” and reality is massive: my 1,138 typed prompts resulted in just over 14,000 distinct model calls (12 per prompt), and each prompt consumed on average 2.9 million tokens. For comparison, typical web-based AI chat exchanges with no reasoning or web searches only use around a thousand tokens.

Over the past the 8 weeks, my Claude Code used 3.2 billion tokens. These overwhelmingly came from the agent re-reading its own working memory. Every time an agent takes a step (e.g. runs a command, reads a file, or calls a tool), the model re-processes its entire accumulated context. The figure below shows the breakdown of how tokens were used and their share of total electricity use.

Token and estimated electricity composition of my Claude Code usage, May 31 to July 25, 2026. “Cache reads” are previously processed context re-read from the key-value cache on each model call; “cache writes” are new context being processed and stored; “output” is text and code generated by the model. Electricity shares use Bistline (2026) factors with cache reads at 10% of the fresh-input energy rate.

The text I actually see (e.g. the model’s output) is only around 0.4% of total tokens processed. Some 96% of the tokens are cache reads where the agent re-reads its own context at each of those 14,000 steps. This matters enormously for the energy estimate, because a cached token is much cheaper to re-read than a fresh one is to process. AI companies charge about 10% of the price for cache reads compared with fresh content, and I use that ratio as my central energy assumption, with 1% and 25% as bounds.2

Since nobody outside of the labs actually knows the true per-token energy of a frontier model (Anthropic has published no per-prompt or per-token figures, something the Watershed paper politely but firmly flags as the field’s biggest data gap), I ran my measured token counts through three independent published methodologies: Watershed’s activity-tier factors, the per-token factors Simon Couch’s estimates derived from Epoch AI’s work, and the claude-carbon tool’s pricing-inferred coefficients.

Estimated electricity consumption for the 3.2 billion tokens I used under three published methodologies: Watershed activity-tier factors under three cache read assumptions, Couch (2026) per-token factors, and claude-carbon per-model coefficients.

Every one of these methodologies gives an answer between roughly 70 and 330 kilowatt-hours over 8 weeks. The estimate is genuinely uncertain, by a factor of ~2 in either direction. But the broader conclusion is not: counting my 1,138 prompts at published per-chat-prompt rates would have suggested about 0.3 kWh, while the reality is 150 to 1,200 times that.

My daily pattern of energy use is shown in the figure below. The day to day variability is huge: my heaviest day (11 kWh central estimate) involved multiple parallel agents churning through a large geospatial analysis, and used more than a third of the total daily electricity of an average US home. This reflects that fact that even within the category of agentic usage, the complexity of the task and the number of simultaneous sub-agents used will greatly influence the resulting energy use.

Estimated daily electricity consumption of my Claude Code usage(bars: cache reads at 10% of input energy; whiskers: 1% to 25%). Reference lines show typical daily electricity use of a refrigerator and of an average US household.

My numbers are a bit higher than some of the other published estimates of agentic use, and it is worth digging in a bit to determine why. Couch estimated that a median Claude Code session uses around 41 Wh, involving 24 model calls and 592k tokens. Andy Masley’s June 2026 calculator puts a 100k-token Claude Opus agent session at ~459 Wh. My median session is ~600 Wh, involving a hundred-plus calls and around ten million tokens including numerous subagents for large data analyses projects. Hannah Ritchie’s hypothetical heavy user (24 agentic queries a day) came out at 2.4 kWh/day, while I measured a central estimate of 3.0 kWh/day (1.2 to 5.9 kWh) for my actual usage.

None of these estimates are necessarily wrong, they just reflect a wide range of actual usage assumptions. Software engineers, researchers, and data analysts (e.g. folks like me) probably lie pretty far down the tail of the usage distribution. At the same time, usage will likely grow over time as more complex agentic tools increasingly become the norm.

What a year of this looks like

If we assume that these 8 weeks are fairly typical, we can estimate that a full year of my agentic Claude Code use would consume roughly 1.1 MWh of data center electricity (0.4 to 2.2 MWh), which is about a tenth of what an average US household uses. Applying the US-average grid intensity, that is roughly 370 kgCO2e per year (150 to 730 kgCO2e).3

Annual emissions of common activities compared with my annualized Claude Code usage. Car: EPA typical passenger vehicle (22.2 mpg, 11,500 mi/yr). EV: 11,500 mi/yr at 0.30 kWh/mi on the California grid. Flight: ICAO-method economy round trip, CO2 only. Home electricity: EIA average US household on the US-average grid. Dryer: typical electric clothes dryer at ~770 kWh/yr (DOE) on the US-average grid.

My personal and professional AI usage now emits a bit more per year than running an electric clothes dryer, and about half as much as driving an electric car 11,500 miles in California or taking one San Francisco to New York round-trip flight in economy.4 It is about 8% of the annual emissions of a typical American gasoline car, and roughly 2% of the average American’s ~18-ton annual greenhouse gas footprint.

This is simultaneously a large emissions source and a relatively modest part of my total carbon footprint. I typically take a round trip flight from San Francisco to the East Coast twice a year to visit my aging parents (not to mention work travel), and I generally don’t lose sleep over that choice. It is also fundamentally a much easier-to-decarbonize end-use than aviation (more on that below). But this also represents a net new source of emissions, at a time when global temperatures are skyrocketing and our emissions reduction goals are increasingly off track.

So what do we do about it?

Having spent most of this post arguing that agentic AI use is hundreds of times more energy intensive than the chatbot numbers suggest, let me be clear that I don’t think the answer is guilt or abstinence. But there are real levers here that we can use to shape the trajectory of AI energy use and emissions going forward.

On the personal side we can try and not be frivolous with agentic tools. There is a real difference between pointing five parallel agents at a hard research problem and doing the same to settle a bar bet (or, in my case, making axolotl-themed games with my daughter). What models you use matters too: sending simple tasks to smaller models uses perhaps 5 to 7 times less energy per token than defaulting to a frontier model,5 and it is what I increasingly do for searches and mechanical work. That said, I don’t want to oversell this. My entire annual AI footprint is a few hundred kilograms of CO2; personal restraint by the small population of heavy users is not going to bend any curves.

The technology lever is more powerful, and it is genuinely impressive. The figure below shows the energy efficiency of NVIDIA’s data center chips over the past decade. The amount of math an AI chip can do per joule of energy has grown roughly 150-fold since 2016, doubling about every two years, per Epoch AI. The latest B300 chips running at their lowest supported precision use about a quarter of the energy per operation of the 2022-era H100s that trained today’s frontier models. This represents a 3.8-fold improvement in energy efficiency in three years. Software gains can make this even faster: Google reports the energy of a median Gemini prompt fell 33-fold in a single year through a combination of better models, better tools, and better hardware.

Peak dense tensor throughput per watt of rated chip power for NVIDIA data center GPUs, by release year and numeric precision. Dashed line shows Epoch AI’s trend of energy efficiency doubling every two years for leading ML hardware.

But if 150-fold efficiency gains were going to reduce AI’s energy use, they would have done it by now. This is the Jevons paradox in action: making compute cheaper per token in turn tends to lead to greater levels of AI use. Efficiency is why my agentic habit costs 170 kWh rather than the 950 kWh it would have used with 2020-era hardware. But efficiency only determines how much intelligence we get per unit of energy, but so far it has so far shown no sign of determining AI’s total energy use.

Which is why the lever that actually matters most is the carbon intensity of the electricity. Every number in this post assumed the US-average grid; run the same workload on largely clean power and my footprint falls by roughly 90%. Unlike aviation, this is an end-use we already know how to decarbonize.

The problem is that we are moving in the wrong direction today: a sizable portion of planned US data center capacity intends to build its own behind-the-meter generation, and nearly three quarters of that is natural gas. AI companies with genuine climate commitments need to do better at finding alternatives: solar plus storage (which I helped lead a study about in 2024), next-generation nuclear and restarts of retired reactors, enhanced geothermal, and siting data centers in regions where both the average and the marginal generation is low-carbon).

There is also a silver linings version of this story where AI demand becomes an asset for decarbonization. Getting to net-zero emissions requires roughly tripling electricity generation by mid-century as we replace nearly all the current uses of fossil fuels with clean electricity. The barriers are mostly not technological, but rather things like interconnection queues, permitting, transmission. The AI buildout is a preview of that world of rapidly increasing electricity demand, backed by companies with enormous capital and unusual urgency. If that money and impatience gets spent speed-running the elimination of those barriers (buying firm clean power, funding transmission, absorbing the early costs of advanced nuclear and geothermal the way early corporate buyers did for wind and solar), the AI boom could leave the grid cleaner than it found it. If it gets spent on behind-the-meter gas turbines, it won’t. That choice is being made right now, and it will matter far more than how many prompts any of us type.

In the interest of full disclosure: the python code underlying the analysis and figures in this post were, naturally, built with the help of Claude Code, but the writing is all mine.6

1 Claude Code records API usage including the amount of uncached input tokens, cache-creation tokens, cache-read tokens, and output tokens, per model call, with model IDs and timestamps. One logging subtlety matters a lot: each API response is written to the log as one line per content block, with every line repeating the message’s full usage object, so a naive line-by-line sum double-counts tokens by a factor of ~2.2. All numbers here count each API message once, deduplicated by message ID.

2 A cache read retrieves already-computed attention states from memory rather than recomputing them, so it is much cheaper than inputting fresh data, but its not free. Cache reads context still makes each output token more expensive to generate at long context. Anthropic prices cache reads at 10% of fresh input, and the claude-carbon and Couch methodologies both adopt ~10% as an energy ratio. Watershed flags cache handling as a known gap in per-token accounting; my 1%-25% band is intended to span the plausible range.

3 Using the eGRID 2024 US-average 341 gCO2e/kWh, since Anthropic does not disclose where their data centers are located and what electricity sources they use. Market-based emissions (counting providers’ clean power purchases) would probably be lower, potentially much lower. This estimate excludes my laptop, which at ~50 W is negligible against 3.0 kWh/day of data center load.

4 Note that the flight estimate here only includes direct CO2 emissions from aviation; including contrails and other secondary factors would probably increase flight emissions by at least 50%.

5 Based on inferring energy use through token pricing, claude-carbon gives ~0.3 J/token for Haiku-class vs ~2 J/token for Opus-class models.

6 In a good example of why you always need to double check work done with AI coding tools, Claude accidentally doubled its original estimate of my token use as all the relevant files were stored twice and it simply added them all up. I only caught it because the numbers seemed too high!

 

Back in 2025, wildlife biologist Kylie Brunette was going through footage from camera traps she’d placed in front of woodcock nests in the forests of West Virginia when something shocking caught...

Ascend Analytics says more than 80% of new large loads seeking interconnection will not have matching generation online by 2030, even as data centers and industrial demand continue to surge.

The New York-based utility company is pursuing growth through electrification, in alignment with city and state climate policies, it said last week.

By Jeremy Brecher,
Senior Strategic Advisor, LNS Co-Founder

Listen to the audio version >>

The startling growth of hyperscale data centers for processing artificial intelligence threatens to exacerbate the climate crisis, the jobs crisis, and the affordability crisis. This commentary presents an overview of the data center explosion and its likely effects. The next commentary in this series will examine the emerging “Data Center Rebellion” and how it can become a major force promoting a Greentech New Deal.

Aerial view of data centers intermingled with other commercial buildings in Loudoun County, near Ashburn, November 26th, 2025. Photo credit: Theodore Christopher, Wikipedia Commons, CC0 1.0 Public Domain.

Data centers are large warehouse-like structures filled with digital electronics that process artificial intelligence (AI). According to the International Energy Agency, “conventional” data centers can use between 10 and 25 megawatts of electricity while “a hyperscale, AI-focused” data center can use 100 megawatts or more. Hyperscale data centers can encompass more than a million square feet. Today there are 4,149 data centers in the US, with 2,788 more under construction or planned. Data center construction rose more than 34 percent between March 2025 and March 2026.

The explosive growth of data centers is already having a significant impact on the energy system. It will have an even greater impact on the transition from climate-destroying fossil fuel energy to climate-protecting Greentech energy in the future. So far, those effects are overwhelmingly negative.

The future of data centers, and AI more broadly, is filled with unknown unknowns. AI is pretty clearly both a technological revolution that will change the way we do things in many spheres of life – but also a technological bubble based on extreme claims that are unproven hype at best. Google CEO Sundar Pichai says artificial intelligence is “more profound than, I dunno, electricity or fire.”

Decisions and investments based on such dubious claims are speculative at best. For example, Chinese AI, with its far cheaper energy system and its cheaper and less energy-intensive strategy, is widely seen as less than a year behind the AI frontier. The AI boom depends on a debt bubble similar to many in the past marked by colossal overinvestment followed by collapse. Tech giants are using other people’s money to make huge investments in hyperscale data centers that have not yet shown they can be even marginally profitable. All of this is developing in the context of Polycrisis 2.0, with its unlimited warfare, unlimited arms races, energy crises, climate crises, galloping inequality, destruction of democracy, and overall prevalence of folly. Finally, the AI bubble is largely driven by the colossal egos of “hyperscalers” like Elon Musk whose megalomania and struggles with each other for dominance will have unpredictable ramifications.

Data centers and jobs

Anti-data-center sentiment in rural Kansas neighborhood, May 24th, 2026. Photo credit: Catboy69, Wikipedia Commons, CC BY 4.0.

AI is clearly affecting employment, but estimates of its employment impact vary wildly. S&P Global’s recent executive survey showed that 42 per cent of organizations abandoned most of their AI initiatives in 2025, compared with 17 per cent in 2024. And a 2024 RAND report indicated that more than 80 per cent of industrial AI projects fail, mainly due to process complexity, poor data quality, and lack of real-world context. The vice president of vehicle hardware engineering at Ford, Charles Poon, explained, “Mistakenly, we thought that by just introducing artificial intelligence and adjusting the design requirements that we had, that that would produce a high-quality product.” Ford recently brought back 350 experienced engineers it had fired and tried to replace with AI.

The building of data centers will undoubtedly create jobs for a sector of construction workers. Nobody knows how many. Like most construction jobs, these jobs are temporary. In a Cologix data center in Columbus, Ohio, construction lasted on average six and a half weeks, with about 146 workers on site at a time. Jobs building data centers often are taken not by local workers but by workers from all over the country who come in temporarily to take the temporary jobs. The Industrial Development Agency in Genessee County, New York anticipates that 60% of the construction workforce at a proposed Stream US Data Centers Project will be from outside of the 14-county region.

Data centers create few permanent jobs. According to Kartik Hosanagar, codirector of the Wharton Business School’s AI research center, “Most data centers employ about one hundred to 200 people. In fact, when Apple created a $1 billion data center in North Carolina, the news stories reported that there were less than a hundred permanent jobs created as a result.”

A study of data centers in Texas found, not surprisingly, that when a data center opens there is a gross increase in data center jobs in a county. But, more surprisingly, these job increases are “offset by job losses in other sectors.” In other words, “though there are gross job flow changes, there is no discernable net change in jobs associated with the data centers in Texas.”

How many jobs will be destroyed by the AI that data centers power? The answer is, nobody knows. Some of the claims made by the AI industry seem extravagant.

“Dario Amodei, the head of Anthropic, has warned that A.I. could eliminate 50 percent of entry-level white-collar jobs within years. The tech investor Vinod Khosla predicted last year that A.I. would replace 80 percent of jobs by 2030. Elon Musk has said the technology will render work ‘optional.'”

Non-industry accounts indicate that the impact of AI on workers will be extensive. For example, researchers at Boston Consulting Group estimated that more than half of the jobs in the United States would be “reshaped” by artificial intelligence over the next two to three years, though far fewer would be replaced entirely. Amazon delivery driver Jonathan Rosenblum gives a vivid description of what that reshaping can mean:

“When I’m in the Amazon truck, every movement I make is tracked with technology and evaluated by AI programs — where I am, which packages I’ve delivered, and whether it’s keeping pace with the algorithm that Amazon has determined I must meet. Readouts at the end of every shift show how each of my deliveries compared to the timing prescribed by Amazon’s algorithmic standard. We are evaluated every week on whether we took accurate photos on delivery, delivered the packages exactly where the customer requested, and got good or bad customer feedback. Through the system, drivers who don’t “make rate” or who don’t meet Amazon’s prescribed standards don’t stay employed.

“Employers everywhere are seeking to imitate the behemoth’s labor model of exploitation, job instability, and — terrifyingly — the deployment of AI technologies to discipline and disempower workers. This workplace dystopia is being perfected at Amazon, then exported to other employers — in factories, grocery stores, hospitals, restaurants, hotels, construction sites, laboratories, and offices.”

AI-related layoff of tech workers at tech firms have made big headlines. But tens of millions of “back office” jobs are also threatened, such as customer service representatives, bookkeepers, payroll clerks, human resources specialists, and many others. These jobs are predominantly, or overwhelming held by women. Says Molly Kinder, a former researcher on AI at the Brookings Institution, “I worry that A.I. will be to high-school-educated women what deindustrialization was to high-school-educated men.

Whatever the future holds, AI is already destroying jobs right now. Researchers at Stanford University found that employment is already declining for entry-level workers in jobs that were highly exposed to A.I. “Early-career workers (ages 22-25) in AI-exposed occupations experienced 16% relative employment declines.”

Derailing the Greentech revolution?

Roof of a data center featuring cooling towers and backup generators, November, 23rd, 2025. Photo credit: Rsparks3, Wikipedia Commons, CC0 1.0 Public Domain.

Data centers use massive amounts of energy. They already used 448 terawatt hours globally in 2025, more electricity than all but 10 countries. That is on track to double within four years. Some data centers consume more energy than a mid-size city. According to Kartik Hosanagar of the Wharton Business School’s AI research center, in some US states data centers are already consuming up to 5% of all energy used in the state. In two to three years that is projected to be over 10% in most states. By 2030, data centers may consume enough electricity to increase the annual growth in electricity demand nearly five-fold. This July, BloombergNEF’s estimate of projected U.S. data centers’ power capacity by 2035 increased 83% compared to their estimate just half a year earlier. Data centers would account for about 20% of total US electricity consumption by 2035, up from less than 6% today.

US utilities are racing to build new fossil-fuel plants and are keeping ageing gas and coal plants open to meet the swelling requirements of data centers. But often the grid simply can’t keep up. Delays are holding up data centers’ connections to the electric grid by as much as 12 years. So big tech is investing massively to produce its own power by any means available. The biggest growth is in the gas industry, including fracking firms and pipeline companies. Gas companies are building new plants solely to supply data centers, leading to the largest ever construction boom of natural gas-fired power plants.

Some of the demand for electricity is being met by Greentech. For example, Google just developed the world’s largest grid-scale battery to power a data center in Minnesota, and purchased an energy company with which it is expanding renewable development, including a new “off the grid” center in Texas that will include wind, solar, batteries, and gas.

But the new energy for data centers is coming overwhelmingly from fossil fuels. Since 2023, some communities have seen a 48% increase in greenhouse gas emissions because of data centers. An example: Mississippi Power agreed to keep burning coal at one of its plants for roughly a decade longer than planned. In Michigan and other states, data centers have effectively derailed the grid’s planned transitions to renewable energy. Meta plans to build 10 gas power plants across Louisiana for its data centers.

The AI industry and data centers represent an extraordinary concentration of wealth and power.

According to the investment firm Jeffries, Amazon, Google, and Microsoft account for more than half of the country’s data center power capacity. Due to the data center boom, Microsoft consumes nearly four times as much electricity as it did before the pandemic; Google’s electricity use has more than doubled. Amazon’s emissions increased by 16% last year; Google’s increased by 18%; Microsoft’s grew by 25%.

Data centers are dramatically increasing electricity costs for utilities’ residential and business customers. Utilities requested $18.6 billion in electric and gas increases in the first six months of this year, according to a new report from the nonprofit PowerLines. More than $9 billion of those requests were filed in the second quarter of 2026 alone, up 26% from the same period in 2025. On July 14 the nation’s largest electrical grid operator said rising electricity costs would add $6.3 billion to the bills of millions of households and businesses within the next three years as a result of the power demands of data centers. Some reports estimate electricity bills may double by 2039 if data centers continue to be built at current rates.

Unconstrained, the rise of hyperscale data centers threatens to accelerate greenhouse gas emissions and thereby put climate change on steroids. It also threatens the jobs of millions of workers, local environments, neighboring communities, and the affordability of electricity, water, and other necessities of life.

The next commentary in this series will describe the “Data Center Rebellion” that is emerging from the grassroots around the country and presenting an unprecedented challenge to hyperscale data centers and the tech oligarchs who are attempting to impose them on American communities.

Get “Strike!” via EmailGet “Strike!” via Substack DONATE ONLINE

The post Will Data Centers Derail the Greentech Revolution? first appeared on Labor Network for Sustainability.

The Virginia State Corporation Commission said it may use an upcoming docket to weigh whether this policy “could or should” also apply to more upstream transmission costs.

In June, a district court judge rejected PSCo’s effort to acquire roughly 550 feet of easement via eminent domain to connect two nearly complete wind farms totaling more than 1 GW. The utility says the project delays are harming ratepayers.

Print article Share

An upcoming UK government consultation on weakening targets for electric vehicles (EVs) could cost consumers as much as £3bn a year by 2030, according to Carbon Brief analysis.

It could require the UK to import an extra 17m barrels of oil in 2030, raising expected net imports by 8%, as well as adding 2.5% to national emissions that year, the analysis shows.

After years of fierce lobbying by parts of the car industry – and despite the significant savings on offer for EV drivers – media reports suggest that EV targets could be “watered down”.

Under current rules, battery EVs – BEVs, those which run only on electricity – must make up a rising share of new car sales in the UK.

This policy, known as the “zero-emission vehicles” (ZEV) mandate, was introduced by the previous Conservative government and sets a goal for 33% BEV sales in 2026, rising to 80% in 2030.

(Carmakers are able to use “flexibilities” to help meet their targets, which reduces the effective target under the ZEV mandate to an estimated 25% of sales in 2026.)

Now, the government under new Labour prime minister Andy Burnham is reported to be considering a cut in the BEV target for 2030 to just 50% of new car sales, alongside options for 60% or 70%.

Carbon Brief understands that a consultation on weakening the ZEV mandate is being reviewed by the prime minister’s office in Number 10, ahead of being formally released.

If the mandate is weakened to 50% by 2030 – and if carmakers make more use of “flexibilities” – there could be up to 3m fewer BEVs on UK roads by 2030, according to the NGO T&E.

Previous Carbon Brief analysis found that BEVs are around £1,100 cheaper to run per year than a petrol car, thanks to far lower fuel costs.

Overall, BEVs are more than £1,000 per year cheaper to own than either petrol cars or plug-in hybrids (PHEVs, which can run on petrol or electricity).

This is according to analysis of the “total cost of ownership” by the Energy and Climate Intelligence Unit (ECIU), including purchase price, fuel costs, insurance and proposed pay-per-mile charges.

In total, Carbon Brief analysis shows that UK drivers could be hit with an extra £3bn in annual ownership costs by 2030, if the ZEV mandate is weakened, as shown below.

A weaker ZEV mandate could “put billions of pounds of committed investments at risk”, reports BusinessGreen, including in the EV charging network and battery supply chains.

Industry group Energy UK says that the mandate is “working in the way it was designed to work” and that it is the “single biggest driver of emissions reductions” in government climate plans.

However, Carbon Brief analysis shows that a weaker ZEV mandate could result in an extra 7.4m tonnes of carbon dioxide emissions (MtCO2) in 2030. This would add the equivalent of 2.5% to national emissions in 2030, under the UK’s international climate goal for that year.

In addition, a weaker ZEV mandate could result in the UK needing to import an extra 17m barrels of oil in 2030, equivalent to 8% of projected net imports that year.

Energy UK says that shifting to EVs will help to reduce household energy bills “for everyone”. This is not only through direct cost-of-ownership savings for EV drivers, but also by spreading the costs of upgrading the electricity system across a wider user base.

Car industry group the Society of Motor Manufacturers and Traders claims that its members are spending “blilions…on discounts, finance incentives and marketing support” and that “natural” EV demand is below the level required to meet the current ZEV mandate. Its claims are disputed.

related Q&A: Does the world need ‘carbon capture and storage’ to reach net-zero? 03.08.2026 Technology UK withdraws millions in funding from world’s second-largest rainforest in Congo  15.07.2026 Nature 28 quotes from new UK leader Andy Burnham on climate, net-zero and fossil fuels 14.07.2026 Policy Analysis: UK newspapers have already printed 63 editorials in 2026 backing North Sea drilling  01.07.2026 Oil and gas

The post Analysis: Weaker EV targets could cost UK consumers £3bn a year by 2030 appeared first on Carbon Brief.

Missourians delivered a powerful endorsement of conservation this month, voting overwhelmingly to renew the Parks, Soils and Water Sales Tax with an unofficial 82% approval. The voters showed...

Ten Senate Democrats are accusing the Interior Department of using a routine-sounding policy review to lay the groundwork for stripping wilderness protections from nearly 200 million acres of public land. In a letter to Secretary Doug Burgum, Sen. Martin Heinrich of New Mexico and nine colleagues wrote that the review “undermines Congress’ role in wilderness management.”

The review, opened through three Federal Register notices in June, covers land managed by the Bureau of Land Management, National Park Service, and Fish and Wildlife Service, including up to 91.4 million acres at BLM alone, according to the Wilderness Society. The public comment period closes August 14.

Heinrich called the process unusual. “It is atypical for the Interior Department to conduct land management planning by asking for comment on its agencies’ general operating procedures,” he wrote, adding that it’s “really just a ploy to weaken protections for public lands,” pointing to the administration’s “obsession with development and disposal of public lands.”

Trump’s “energy dominance” agenda is only helping oil and gas companies

President Donald Trump promised to “cut your energy prices in half” within his first year in office. Instead, “energy dominance” has produced record profits for oil companies and higher gas prices for everyone else. Trump has used his power to cut the cost of drilling on public lands, while driving up the price of oil by waging war on Iran. Eight of the world’s largest oil companies made a combined $93 billion in profit in the second quarter of 2026, nearly double what those companies made last year and more than $1 billion in profit every single day. Read more in a new Westwise blog post on Substack by CWP’s Lilly Bock-Brownstein and Rachael Hamby.

Quick hits People are betting millions to predict wildfires. Congress wants to stop it

Outside

Opinion: Sportsmen should get a say on our public lands, that’s why we need the Local Input Act

Colorado Times Recorder

Bulldozing begins at Big Bend National Park for border wall construction

Washington Post

New Mexico delegation introduces bill to ban uranium mining in Chama Watershed

Source New Mexico

An Indigenous photographic history of America in pictures

The Guardian

White House asks Justice department to consider new reflecting-pool prosecution

Wall Street Journal

House Democrats seek records in dropped Reflecting Pool case against ex-Olympian

Washington Post

The Reflecting Pool haunts Doug Burgum

Axios DC

Quote of the day

The BLM used to be able to look at a nominated parcel, listen to the people who use it, and hold it back when leasing would, for example, impact elk or mule deer calving grounds, choke off a trailhead, or industrialize high-priority habitat. That discretion is gone. The oil company picks the parcel; the BLM is told to process the paperwork. End of story.”

—David A. Lien of Colorado Springs, Colorado Times Recorder

Picture This

@Interior

The monarch migration is about to begin. Along the way, they depend on native flowering plants to rest and refuel. Planting native wildflowers and milkweed in your yard or in your community is one simple way to help these remarkable travelers complete their journey.

Feature image: Rainbow Mountain Wilderness: Source: BLM Nevada, by Bob Wick/Flickr

The post Democrats call Interior wilderness policy review a ‘ploy’ appeared first on Center for Western Priorities.

A new draft text to revive deadlocked UN plastics treaty talks does not include specific measures on managing runaway plastic production, a growing source of greenhouse gas emissions, drawing criticism from some countries and campaigners that ambition for the global pact is shrinking.

After diplomats met in Nairobi early in July for the first time since negotiations fell apart a year ago, Chilean ambassador Julio Cordano, who is chairing the talks, released a first document last weekend, setting out elements of a possible treaty to tackle plastic pollution.

Cordano stressed this is an “informal reference document” rather than a negotiated text. But its structure is similar to a draft treaty and closely resembles the previous version rejected by governments during the last round of formal negotiations in Geneva.

The new text recognises the world’s “unsustainable” levels of plastic production and consumption, both of which are projected to nearly triple by 2060. But it contains no measures to stem that growth, critics say, pointing to what they see as a broader weakening of ambition.

They argue the document is increasingly aligned with the demands of fossil fuel-producing countries, including Gulf states, the US and Russia, which have pushed for the treaty to focus on managing plastic waste rather than limiting production.

“When you leave the countries that have the most vested interests in delaying meaningful action to shape the agenda, you end up with a text that does nothing to end plastic pollution,” said David Azoulay, environmental health programme director at the Center for International Environmental Law (CIEL).

France disappointed with production omission

“We’ve gone backwards rather than forwards,” Christina Dixon, a campaigner at the Environmental Investigation Agency (EIA), told Climate Home News. “A text that was rejected by the majority of countries in Geneva as being too weak and not ambitious enough has been repackaged one year later with some key elements removed and put out as a kind of sign of progress.”

A French diplomatic source told Climate Home News it was “disappointing” that the text lacked any concrete provisions on tackling “unsustainable” levels of plastics production and consumption. That is despite a majority of countries repeatedly advocating for curbs and scientists saying the world cannot put an end to plastic pollution without tackling the issue at source, they added.

• Aug 11, 2026 Oceans South Africa’s offshore oil push meets grassroots resistance in court Two court cases against oil majors TotalEnergies and Shell could derail South Africa’s plan to expand offshore fossil fuels Read more
• Jul 29, 2026 Politics UN bid to keep 1.5C alive exposes deepening divisions over fossil fuels Vulnerable island nations and the poorest countries want the Belém Mission to 1.5 to drive fossil fuel phase-out, while Arab states call for more investment in oil and gas Read more

Governments across Europe, Latin America, Africa and the Pacific islands have previously called for efforts to limit the manufacturing of plastics to “sustainable levels”, but their efforts have been frustrated by strong and persistent opposition from a small group of fossil fuel producers, who see plastics as a growing market for oil and gas.

Weakening of production ambition

Cordano told Climate Home News that the “concept” of sustainable production is still reflected in different parts of the new document.

But measures aimed at achieving that objective have progressively weakened over time. Initial versions of the draft treaty, dating back to 2024, included a standalone article with the option of setting a global target to reduce the production and consumption of primary plastics.

That disappeared from successive drafts published in Geneva last year. The last version nevertheless said data on plastic production could be considered in future assessments of whether the treaty was meeting its objectives. Observers saw this as an important provision that could have strengthened the pact over time and potentially kept the door open for a global production target.

The new text only mentions “sustainable production” in the preamble and includes an article saying that countries could improve the design of plastic products in order to contribute to “sustainable production”.

“There’s a war of attrition element,” said Dennis Clare, a negotiator for the Pacific island nation of Micronesia. “The countries that want to do less are dragging out discussions and gradually pressuring the more ambitious to compromise towards a lower common denominator.”

Little space for thorny discussions

Countries have twice failed to agree on a global plastics treaty at what were meant to be final rounds of negotiations in December 2024 and August 2025. After being selected as the new chair earlier this year, Cordano has been working to steer the process back on track through a series of informal meetings, hoping diplomats can find common ground ahead of the next formal negotiations scheduled for early 2027.

But he has been criticised for sidelining discussions on some of the thorniest issues. Cordano kept plastic production off the official agenda for the Nairobi meeting a few weeks ago. He said beforehand that countries could bring any issue to the table, but production did not feature in the summary of discussions subsequently published by the chair.

Clare said discussions on fundamental elements of the treaty, including production, had been “constrained” and that there was little space for them in Nairobi.

Cordano told Climate Home News the Nairobi talks had provided space both for “reaffirming positions and expressing new ideas”, adding that countries “remain free to raise all issues they consider important”.

Informal talks between negotiators are held behind closed doors and neither the media nor external observers can take part.

Workers sort plastic waste at a recycling workshop on November 17, 2025 at Xa Cau village, outside Hanoi, Vietnam. (Photo by Thanh Hue/Getty Images) Workers sort plastic waste at a recycling workshop on November 17, 2025 at Xa Cau village, outside Hanoi, Vietnam. (Photo by Thanh Hue/Getty Images)

Campaigners have accused the chair of making political calculations to reach an agreement at any cost. “He has clearly identified that the only way to achieve an agreement by consensus is to do away with the more complex elements of the treaty like those that deal with sustainable production and consumption of plastics,” the EIA’s Dixon said.

Cordano said he continues to be guided by countries as “they develop their own exchanges and continue working towards possible landing zones”.

Push for more ambition

Governments will debate the new text at another meeting of chief negotiators in Bangkok, Thailand, at the end of September, and a new version of the document is expected after that meeting.

The French diplomatic source said the current text should not be viewed as “an end-product”, but as a starting point that “can and should be improved”.

France, together with the EU and members of the High Ambition Coalition (HAC), will continue pushing for stronger provisions, including measures to address plastic production, the source said.

China’s coal power rebounds as record clean energy goes to waste

The HAC group includes over 70 countries, primarily from across Europe, Latin America, Africa and the Pacific.

Micronesian negotiator Clare said countries on the frontline of the plastics crisis may decide to reject a really weak treaty that puts the burden on them to clean up somebody else’s waste, while producers can keep churning out plastics unrestrained.

“If the treaty does not include essential elements of the solution, even an initial, apparent diplomatic success – an agreement – can come to be seen over time as an environmental failure,” Clare warned.

The post “We’ve gone backwards” – new plastics treaty text dims hopes for production curbs appeared first on Climate Home News.

CIW signs an agreement with the Florida Tomato Growers Exchange (2010), leading to the creation of the Fair Food Program ProPublica: “(T)he gold standard for preventing worker abuse is the Fair Food Program”

Last week, we shared with you the news that the largest modern-day slavery prosecution in U.S. history, dubbed “Operation Blooming Onion,” had just wrapped up. In updating you with the latest developments in that case, which the CIW uncovered over a decade ago and helped federal authorities investigate, we also reflected on the importance of preventing the kind of abuses that occurred within the Blooming Onion operation — forced labor, sexual violence, wage theft, threats, and retaliation — before they ever have a chance to take root.

This week, for our final post in this year’s Sustainer Drive, we want to share a feature-length piece from ProPublica that approaches the problem from the perspective of prevention, and to let you know how you can help be a part of the solution.

After reporting extensively on Operation Blooming Onion and the workers who suffered under that sprawling criminal conspiracy, ProPublica reporter Max Blau set out to answer a deceptively simple question: How can any of us know whether the fruits and vegetables in our grocery carts were harvested by workers whose fundamental rights were protected?

His search led him to the Fair Food Program — a program that has demonstrated how the kinds of abuses in Blooming Onion can not only be remedied after the fact, but prevented in the first place. Below are some excerpts from the ProPublica story:

My Maddening Quest to Find Out if Blueberries Come From Farms Where Workers Were Harmed

“Over the past year, I’ve reported at length about crimes against farmworkers who have come to the U.S. legally from Mexico and Central America.

I haven’t stopped thinking about how they suffered. Their wages were stolen, guns were pointed in their faces, and one woman I met was held captive, raped and nearly killed.

As I wrote those stories, there was something else I couldn’t stop thinking about — something I keep thinking about: Whenever I go to the grocery store, I have no easy way of knowing if the fruits and vegetables I put in my cart are part of the problem I’ve documented…

That case, which I wrote about at length, revealed that thousands of foreign farmworkers had been illegally charged fees by labor contractors to work in the U.S. Some were forced to pick crops for little to no pay in what prosecutors described as a form of modern-day slavery.

This wasn’t the first case of the abuse of farmworkers at the hands of labor contractors. Or the last. Many of these harms might have been prevented if the workers had received the protections they were promised by the U.S. government. And there have been no substantial reforms to address the abuses, either…

I spoke with experts to see if there are ways for consumers to know if the people picking their blueberries are being mistreated.

Those experts, who study farm labor conditions and food supply chains, said the gold standard for preventing worker abuse is the Fair Food Program. Under the program, supermarkets and other corporate buyers commit to paying a little more for produce — a cost so negligible, at a few pennies per pound, that one expert said it shouldn’t drive up prices.

For those few pennies, grocers get crops from farms that offer their workers stronger protections.

Fruits and vegetables from those farms are labeled with a green sticker of a woman holding a basket of tomatoes. The stickers give consumers like me a clear indication that the workers who picked that produce were treated fairly.

When I’ve found these tomatoes in my grocery stores, they don’t cost measurably more than others without the label.

Dozens of farms have signed on to the Fair Food Program, and participation in it has been shown to prevent abuse in the fields.

But only a few small blueberry farms participate in the program.

Until that or something else changes, consumers will know virtually nothing about the farms their blueberries come from. And they won’t be able to make choices that could make a difference in the lives of the people who pick them.”

The question ProPublica leaves us with is the same one that has driven the CIW’s work since the launch of the Fair Food Program nearly two decades ago:

Now that we know how to prevent these abuses, how do we make those protections available to every worker who needs them?

The Fair Food Program has already demonstrated what prevention looks like in practice. But for true prevention to become a reality across U.S. agriculture, the Program must grow dramatically — reaching countless more farming operations, crops, and communities, and ultimately protecting hundreds of thousands more workers every year.

The need to scale is only becoming more urgent by the day. Outside the protections of the Fair Food Program, workers remain vulnerable to the very abuses documented in Operation Blooming Onion. To make matters worse, the growing pressure on the country’s existing immigrant worker community is driving more and more growers across the country to turn to the H-2A “guestworker” visa program to recruit farmworkers, a program with a long and documented history of labor abuse and the same program the Blooming Onion ringleaders exploited to control and mistreat thousands of workers over nearly a decade. As the risks facing workers multiply, the life-saving protections of the Fair Food Program — protections created by workers themselves and backed by real, enforceable consequences — will only become more essential.

And that is why your support matters so much.

Monthly gifts help ensure that the FFP’s protections can reach more workers, on more farms, in yet more distant corners of the agricultural industry from Florida to California.

Today, the Fair Food Program is like the earliest days of the electrical grid in this country. In those early days, the electric grid, in the rare community where it existed, worked beautifully and transformed everyone’s lives, from great industries to the smallest households. But it only covered a small percentage of the entire population, and great swaths of the country remained in the dark. With time, significant resources, and great effort, however, a mass electrification movement was mounted to bring the miracle of electric light to homes across the country, and a social and economic revolution ensued.

The Fair Food Program today covers tens of thousands of farmworkers in 23 states, which sounds great… until you realize that those tens of thousands are only a small part of a population estimated to be more than one million across the country. And, yes, just like electricity, where the FFP operates it transforms everyone’s lives — from the individual farmworker to the grower and even the billion-dollar retail food giant alike. Indeed, like the electrical grid, the Fair Food Program is a shared infrastructure that operates in the background to meet everyone’s needs, ending labor abuse for workers, eliminating regulatory and reputational risk for growers and buyers, and providing ethically grown and harvested produce for consumers. It is a true miracle, one that transformed fields that federal prosecutors famously dubbed “Ground Zero for Modern-day Slavery” into what human rights experts called “The best workplace environment in American agriculture” on the front page of the New York Times in the space of just three years.

But as it was with electricity in the early days of the last century, the challenge before us today isn’t inventing the FFP; it’s building its protections out so that its life-changing power can be experienced by everyone, on every farm, big or small, in every corner of the country. We stand on the doorstep of great change, but we are not there yet.

So, as we close this year’s Sustainer Drive, we are asking you to help us take the next step. If you are not yet a Sustainer, become one today. If you already give monthly, consider increasing your sustaining gift.

The solution to human rights abuse exists. Together, we can make sure it reaches the hundreds of thousands of workers who still need it.

Together, we can complete the human rights revolution in the fields that started in Immokalee 30 years ago this year.

Become a Sustainer today.

Today’s Consumer Price Index (CPI) report shows prices rose 0.1% in July and are up 3.4% over the past year. July’s price increase reverses the disinflation seen in June, which was built on a short-lived ceasefire that quickly unraveled. Nearly six months into his war with Iran, the president has failed to swiftly end the conflict and recently said he does not “think about Americans’ financial situation” when negotiating an end to the war he started.

Meanwhile, oil prices are back on the rise, gas is back above $4.00 a gallon, and Trump’s latest tariffs on 99% of imports are now in effect. As the war drags on and Trump’s tariffs work their way to store shelves, prices are poised to reaccelerate in the months ahead. This price squeeze is happening at the same time that workers are struggling to find jobs. The economy lost 23,000 jobs last month, while wages failed to keep pace with prices. Prices are up, jobs are down, and working families are bearing the brunt of it all.

Groundwork’s Senior Vice President of Policy and Advocacy, Alex Jacquez, released the following statement:

“Prices started climbing again in July and Trump’s catastrophic mismanagement of our economy means more spikes in the months ahead. The president said it himself: he is only ‘semi-negotiating’ an end to the war, all while slapping new tariffs on 99% of our imported goods. Trump is not serious about bringing much needed relief to working families who are weary of higher prices at the pump and on the shelves, and who are pessimistic about an economy that is bleeding jobs.”

To speak to a Groundwork expert about this month’s CPI reading, email press@groundworkcollaborative.org.

BACKGROUND

Trump’s war is keeping gas prices high, while families face higher electricity prices to keep cool.

• Energy prices are 14.7% higher than one year ago, with gasoline up 24.6%. Airfares rose 2.2% in July as jet fuel prices climbed, and are up 25.5% over the past year.
• Prices at the pump have climbed above $4.00 per gallon since the collapse of Trump’s ceasefire, and remain 35% above their pre-war level, turning routine trips to work, school, the grocery store, and the doctor’s office into a bigger strain on household budgets. The average household has paid an extra $634 in gas and diesel expenses because of Trump’s war.
• Keeping the lights and AC on this summer has cost families more. Electricity prices are 4.2% higher than one year ago. The average family will spend $792 on their summer cooling bills, up 10.5% from last year. This comes at a time when one in six households is already behind on utility bills.

Paychecks are losing ground to higher prices.

• Wages rose just 3.2% over the past year in July – the slowest pace since 2021 – while prices rose 3.4% over the same period.
• Real private-sector wages fell 0.4% over the past year in the second quarter of 2026, and the savings rate has dropped to 2.7%, the lowest since 2022. Workers’ purchasing power has weakened, and the cushion they lean on during hard times is at risk of disappearing.

Families continue to face higher prices at the grocery store.

• Trump’s war with Iran is increasing the cost of moving food across the country. Diesel prices are up 46% since the war began, raising transportation costs throughout the food supply chain, particularly for fresh and refrigerated products that depend heavily on trucking. Those higher costs are passed along to shoppers at the grocery store, where prices for fresh fruits are up 2.2%, baby food 0.7%, and milk 0.5% in July alone.
• Trump’s tariffs are also showing up in the grocery store. Prices for primarily imported foods such as coffee and seafood increased 10.3% and 7%, respectively, over the past year.
• Food prices remain a major source of financial strain for families. A new McKinsey Institute for Economic Mobility report found that 90% of respondents cited groceries and food prices as a top concern.

Trump’s tariffs are making everyday household goods more expensive.

• Prices for tariff-exposed goods continued to rise in July. Dishes and flatware are up 10.4% over the past year, cookware and tableware up 6.4%, sporting goods up 4.3%, and apparel up 3.9%. Today’s CPI report captures only one week of Trump’s newest tariffs on 99% of imports, and does not yet show much of the impact of Trump’s new pharmaceutical tariffs, which began to take effect July 31.
• Executives are telling investors that tariffs have led to price increases this year, and more price increases may be on the horizon.
  • Whirlpool, a manufacturer of home appliances, told investors it had raised prices twice this year in response to tariffs.
  • Newell Brands, the owner of famous brands like Rubbermaid, Sharpie, and CrockPot, said tariffs would add $127 million to its costs this year, and that previous price increases “don’t come anywhere close” to covering the added costs, signaling the company could raise prices again.

An ongoing drought in Europe has sapped French rivers. Officials say that 43 percent of small waterways in France have been partially or completely dried up.

Read more on E360 →