You don’t need to say it; I’m aware that Amazon has a vested interest in the expansion of data centers. 
That does not automatically invalidate the legitimacy of the arguments presented. The arguments should be assessed on their accuracy, not on who wrote them.
Thoughts on the merits of the arguments?
I used Claude to fact check. This article. However, I haven’t fact checked Claude!
The essay is by Matt Garman, CEO of AWS, published on About Amazon on October 2, 2026.
The essay is an advocacy document, not a neutral briefing. Two parts of it deserve different treatment:
Overall confidence: 75%.
Misleading. Confidence: 80%.
The capital is largely private, but states subsidize data centers through tax incentives:
Ratepayer-borne grid costs (claim 5) are a further public cost outside the tax system. The interstate-highway analogy also breaks down here: highways were publicly governed assets, while data centers are private ones.
Broadly accurate. Confidence: 80%.
As of July 2026, more than 100 local municipalities had adopted or considered data center moratoriums, and at least 14 states had introduced or considered statewide moratoriums or similar restrictions. On July 14, Governor Hochul signed an executive order establishing the first statewide moratorium on new large-scale data centers.
Two qualifications:
Partly supported, rhetorically overextended. Confidence: 70%.
There is real evidence of foreign activity:
The same reporting shows the opposition is large and homegrown:
Gizmodo notes the claim echoes one President Donald Trump has also repeatedly made without evidence.
Accurate on direct use, materially incomplete. Confidence: 80%.
The figures cover only on-site water. They leave out the water used to generate the electricity data centers consume:
Two further gaps:
Misleading; the weakest section. Confidence: 80%.
PJM is the largest regional U.S. grid market. Its independent market monitor and outside analysts contradict the essay:
Garman’s points about aging grids are fair, and I did not verify his state-by-state examples. But selective state comparisons don’t rebut the PJM evidence. His pledge to cover Amazon’s own infrastructure costs also doesn’t address capacity-price spillover, which raises costs for all customers.
Misleading as a generalization. Confidence: 75%.
Virginia’s rules and permits allow far more run time than the essay implies:
Plausible but unverified. Confidence: 50%.
Self-reported. Confidence: 50%.
These are Amazon’s own metrics. The new annual reporting pledge is the right way to test them, if Amazon discloses its methodology.
Accurate as announced, modest in scale. Confidence: 85%.
Not verified:
Secondary sourcing: Several PJM figures come from secondary summaries rather than primary monitor reports.
Timing: Repeating this search in about a week should capture fuller responses from environmental and consumer groups.
Tip: Next time, a one-line instruction with the URL, such as “fact-check this,” gets you straight to the work.
Ridiculous thread. The modern age. An advocacy article from Amazon, not noted for its deep commitment to our wellbeing. A silly vetting of the article about AI by an AI. We are losing our brains, bit by bit.
Katie
Brief summary from Semafor here:
I just want this race to be won on sound economics. Make a profitable product and then build out from there. It’s why I am most interested in Federico’s experimentation with Local AI. The data center backlash is mostly in part that we all acknowledge that our Tech Lords are maladjusted sociopaths. As a millennial that survived 2008, the Covid Economy, and now whatever is going on in the current events that I will only allude to to not get too political, I am tired of obvious bubbles careening us off a cliff and its normal people holding the bill.
Be Transparent with Local Economies. No more midnight deals.
Do not ask for ONE DIME of tax money that is unearned. (Considering how in the toilet their reputation is that might even include subsidizing infrastructure improvements).
Give back to the communities you build them in. Sponsor the local football team for like a decade.
Then you might earn back our goodwill.
The author must have used AI to generate that image of an Amazon employee who looks like he’s enjoying his job and being paid a fair wage.
Personally I wouldn’t believe a word this article says. YMMV.
I always believe in “follow the money” to judge the quality of any statement. In this case you don’t have to look any further than the author. Look for statements about the AI Race from other than vested interests. This could easily be a race that if you win, you lose. See also the Japan’s Fifth Generation Computer “Race” back in the 80’s that also frightened everybody.
In my opinion, AI is ruining this discussion forum.
Sort of. I cringe when we get a 1,000-word post with Claude or OpenAI opinion about an article or whatever. This is not the only forum with that tic. It’s a phase, of course. We all know how to ask AI questions, and we don’t need the answers copied here.
Katie
Interesting. Amazon says data centers bring large numbers of good jobs to a community.
How? AI is essentially cloud infrastructure. AWS data centers fuel a lot of employment, but comparatively little where the data centers are located. I live in a little rural community. My nearest coworker is close to a thousand miles away. All the systems and services I maintain are hosted, um, somewhere. Not here, even though my job depends on them.
Claude did a good job analyzing that article. But what would that analysis have cost if Anthropic hadn’t lost $40 billion investor dollars last year?
I think we’re headed for disaster from reckless investment. We could win the AI race only to lose it through foreclosure.
Oracle stock, staggering a little stronger lately, down 50% over the last 12 months. Fermi and Project Matador in Texas, at 13 cents on the dollar compared to one year ago. NVIDIA’s doing great, but look at how many GPUs they have sold to be shelved and counted as non-depreciating assets.
Be careful out there.
Permanent jobs? From the unofficial comments and stories I’ve read a Google datacenter can employ as few as 50 people to as many as 500 - 1000 at the largest multiple building sites. But we would probably need to talk to the locals to get the real story. IMO
They should be a start, not an end result.
Nothing wrrong with posting it and saying - “I agree with X but I disagree with Y.”
Also nothing wrong with “Claude says this - what errors does anyone see?”
A data center could be colocated with business offices, of course. I think in general developers are remote, often working from home (raising hand), and on site data center staff would mostly be security.
Attracting developers to a site fifty miles from a Starbucks might be difficult. Find Hubbard, Texas on a map. There’s a 600 MW data center being built there for Anthropic. There are only 33 businesses listed on their Chamber’s web site and that counts Nexus, the company building the DC.
Mea culpa. I posted it with a bit of an ironic smirk. The BS of the Amazon article was obvious but I couldn’t let the naive post pass. Sorry I wasted everyone’s time.
I’m familiar with the area, I spent some time near Killeen in the ‘70s. I think that was the last time I saw the Milky Way. No big city lights.
I can understand the problem someone might have recruiting workers.
Not a waste of time here! It was good food for thought, which is an attribute entirely separate from cordial agreement or disagreement.
Thank you for posting the link.