/Tech7h ago

Kim Isenberg and Peter Wang argue the US is losing the AI race due to energy and data center constraints

Story Overview

Analysts Kim Isenberg and Peter Wang frame America's AI lag as a direct result of stalled data center construction and electricity grid limits, contrasting this with China's rapid rollout of models that favor domestic Huawei chips and infrastructure.

1601.3K179629217.5K

#121

Original post

Chubby♨️@kimmonismus#1360inTech

The worst-case scenario for the United States is becoming increasingly realistic, and I will briefly explain why.

@quxiaoyin raised many valid points, and I agree with her. First of all:

-China certainly does not place such strong emphasis on open source because it cares so deeply about humanism, but because it is a strategy to attract many users, gain market share, put pressure on US models, and also because the models are increasingly being trained on Huawei hardware (think of DeepSeek 4), allowing China to host the entire stack domestically.

-But the underlying logic is far more important: The United States is still building too few data centers to meet future demand. @ChrisGillett wrote an outstanding analysis on this, which I shared a week ago. In short, based on SemiAnalysis data, demand is greater than what is currently being built in terms of data centers.

-Even more importantly, however, the United States lacks sufficient energy and grid capacity. This is a problem that will become much more severe in the near future. China, by contrast, is addressing the issue through a massive expansion of its energy supply. Solar capacity: in 2025 alone, China installed as much solar capacity as the United States did in 10 to 15 years. China is also building 36 nuclear power plants, significantly more than the United States, and is installing them faster.

-In addition, China is managing to become more independent through Huawei chips, even though the country still lags far behind NVIDIA. But here, China is betting on quantity rather than quality.

In short: China is a real threat in the AI race, and the situation for the United States is becoming increasingly precarious. This is also the main reason why China is to be kept away from SOTA LLMs at all costs, so as not to jeopardize the lead under any circumstances.

Xiaoyin Qu@quxiaoyin

The worst case scenario for USA AI: 1. Chinese open sources keep gaining market share. China owns the model layer. 2. Those models were trained and inference-optimized on Huawei chips instead of NVIDIA. China also owns the chip layer. 3. US doesn't build data centers fast enough to keep up with the demand of compute, storage and energy. China meanwhile exports the inference and training layer(for continual training it will happen along with inference) Export control is not the right strategy here. Simply banning "open source from China" doesn't solve the issue here. USA must invest in open source models, hopefully get Chinese models to use NVIDIA, and invest in nuclear asap.

2:21 AM · Jun 29, 2026 · 178.4K Views

Infrastructure Lag

Power and build constraints create clear bottlenecks

US data center projects saw roughly half the prior quarter's new starts by late 2025 amid grid limits and local resistance, with forecasts showing potential shortfalls reaching dozens of gigawatts by 2028 even as demand climbs.

Hardware Push

Chinese releases tie models to local hardware

DeepSeek's April 2026 V4 preview, offered under an MIT license with steep early discounts, was optimized in part for Huawei Ascend chips and triggered a scramble among Chinese firms for that hardware, though the weights released remain opaque rather than fully transparent.

Sentiment

Positive users praise China's AI infrastructure buildout and open-source gains as effective, while negative users fault US leaders, permitting failures, and export controls for ceding ground.

Pos

50.0%

Neg

50.0%

15 comments with sentiment.

Cluster Engagement

Digg Deeper

No Digg Deeper questions have been answered for this story yet.

Posts from X

Most Activity

VIEWS38.5KBOOKMARKS43LIKES142RETWEETS14REPLIES26

Garry Tan@garrytan

Build power and datacenters

Chubby♨️@kimmonismus

The worst-case scenario for the United States is becoming increasingly realistic, and I will briefly explain why.

@quxiaoyin raised many valid points, and I agree with her. First of all:

-In addition, China is managing to become more independent through Huawei chips, even though the country still lags far behind NVIDIA. But here, China is betting on quantity rather than quality.

2h38.5K14243

Nikita Belokopytov@NikiBelokopytov

@kimmonismus @quxiaoyin That's why the US and the EU will likely turn to regulations instead of continuing to compete

7h1.5K91

Raph. H.@Rapahelz

@kimmonismus @quxiaoyin American AI companies: bet on closed moats and pricing power → massive capex and temporary lead as a byproduct

Chinese AI companies: bet on openness → massive adoption and ecosystem lock-in as a byproduct

2h723

Cloverx@Cloverx8888

@kimmonismus @quxiaoyin how do you banned "open source"? once its out you cant ban it

7h1.1K1

Chubby♨️@kimmonismus

@Cloverx8888 @quxiaoyin How do you even ban open source? I’d say it’s almost impossible to ban open source

7h99631

Nikita Belokopytov@NikiBelokopytov

@kimmonismus @quxiaoyin I think that the US will push through regulations on business use of models with no safety certificates in the EU, therefore restricting the use of open models. This will allow the US companies to freely expand with no competition from China.

7h3207

Yume_X@yume_arasaki

GLM 5.2 is the real deep seek moment, a sota level model that runs on some measures as well as opus, and gpt 5.5 , design that runs on local hardware easily, cheaper, hackable for a mac studio 512gb or 4x dgx spark (with switch)

The next 6 months will probably just be the rest of the world finally coming to terms with that US capex on frontier lab APIs is basically unsustainable vs China's model.

China is not going to have the smartest models or the best chips, but they don't need to, unlike the smart phone revolution, there is no brand loyalty here. It's straight to whichever AI just gets the job done.

So in analogy, there is no iphone or Tesla here. Apple and Tesla had unique moat in terms of vertical integration, and a luxury brand status. Claude and ChatGPT is never going to be a "flex"

In that way Elon remains to have some of the most secure moats in comparison to other US AI companies.

hardware US companies will be fine. tech companies with software everyone uses e.g. facebook, instagram are fine.

The US AI frontier labs will just get absorbed into big tech companies, and become research division probably, their revenue model makes no sense in the long run, on multiple angles, the only thing driving it is money printed from the top, and US's unique position of controlling world finance (The USD and power of the US central bank and financial plumbing), but this doesn't stretch forever.

Eventually the 2007 of AI will happen.

6h12331

Stella Biderman@BlancheMinerva

@pwang Why should the interpretability of a system matter to whether it’s open source? Nobody’s definition of “open source” “free software” or similar concepts say anything about it as far as I’m aware.

Peter Wang 🦋@pwang

STOP CALLING THEM OPEN SOURCE MODELS

Chinese models are free to use, opaque weight models. They’re less interpretable (and thus less open) than a proprietary .exe file.

But the gist of the tweet is right: the AI race is ultimately an energy race, and the US is behind.

3h27260

Chubby♨️@kimmonismus

@NikiBelokopytov @quxiaoyin What do you mean by that?

7h1.4K1

Klar Moehre@Moehrenmann

@NikiBelokopytov @kimmonismus @quxiaoyin Yes, but that's the most stupid strategy in such a situation. They'll lose even more market share because regulation is a non-market-conform measure. So they'll regulate harder and lose even more market shares.

What bureaucrats never understand are markets and humans.

7h2865

marine_p@marinep655097

@kimmonismus @quxiaoyin Hell yeah, either China wins or genocidal pedophilic capitalist Epstein class wins.

Pulling for China

7h2653

Yair | principal engineer / founder at night@Yairtech

@kimmonismus @quxiaoyin *EU enters the chat: “oh guys we can’t build a datacenter because our rules don’t allow it because we want degrowth so save the planet”

7h5122

Paulo Santos@ThinkFinance999

@kimmonismus @quxiaoyin Energy is a fake problem, and Chinese suppliers would gladly supply the US with as much capacity (solar and wind, as well as grid assets except for the most advanced - which China doesn't yet produce but likely will, soon).

6h6051

Chubby♨️@kimmonismus

@ThinkFinance999 @quxiaoyin energy is a fake problem? What do you even mean by that?

5h5385

Raph. H.@Rapahelz

@garrytan @paulg The central question revolves around whether Artificial GPU Intelligence + Artificial Gigawatt Intelligence → Artificial General Intelligence?

2h1211

Joady@Joadys

A lot of good points. Just one that I disagree with, which is that I don't believe that any of the Chinese models are training on Huawei. I think they're doing some inference on it but I don't think they're remotely good enough. I think them claiming to have been trained on Huawei is a pr stunt because it's important to the government.

6h6711

Simply Complex Systems@EricPedersen81

@kimmonismus @quxiaoyin All data centers will be in space.

They are all coming to terms with that.

xAI will be the only frontier lab by the end of the year and no one will ever catch up.

7h3111

Amx57 † 🌎 ⚛️@retroamxev

@kimmonismus @quxiaoyin We knew the power issue would be a limiting factor....and here we are.

6h2821

Chubby♨️@kimmonismus

@Joadys @quxiaoyin afaik they did some training of DeepSeek 4 on huawei chips, if i remember correctly what i read in their tech paper

5h550

Peter Wang 🦋@pwang

One of the foundational principles for the original open source movement was the right to inspect what’s actually running on your computer. The 1970s way to manifest this was the require the availability of the source code (precursor for executable binaries).

For an LLM, that transparency and legibility has to manifest as interpretability. Indeed, no model is there yet, but that’s what the original spirit of “open source” would require.

1h701