Top AI data center Providers
Four hyperscale cloud providers
Currently, there are four major cloud providers in the United States: AWS, Azure, GCP, and OCI. Investors must closely monitor the movements of these four companies, as they are indicative of future trends.
Tech behemoths
In addition to the four major cloud providers—AWS, Azure, GCP, and OCI—Meta, Apple, and OpenAI, while possessing their own large cloud centers, still rely on other large cloud data providers.
Customer of 4 hyperscale cloud providers
The following is a list of major, well-known customers of the top four hyperscale cloud providers, either publicly or as revealed by the media:
- OCI: OpenAI, xAI, Meta, Nvidia, AMD, Palantir, xAI, TikTok, Temu.
- Azure: Meta, OpenAI, Palantir, Coca-Cola, Siemens, Synthetaic, Lockheed Martin, Raytheon, and UBS.
- AWS: Apple, Workday, TSMC, Veeva, Philips, Grab, Sentinel One, NXP, Schlumberger, Trip.com.
- GCP: Apple, Meta, OpenAI, Salesforce, ServiceNow, X (Twitter).
Current Trends
Enterprises bet on multiple providers simultaneously
Modern enterprises are finding it almost impossible to break away from using data centers. Most companies use two of the four major cloud providers, AWS, Azure, GCP, and OCI, or one of the four plus a smaller provider. The reason is simple: if one provider experiences a service outage, an alternative service must be immediately available to avoid disruption.
4 hyperscale providers rely on outsourcing providers
The demand for computing power and data center infrastructure driven by AI development seems to be on the rise, at least for now. The associated capital expenditures are astronomical, beyond the reach of even the deep-pocketed Big Four cloud computing and hyperscale tech companies. Furthermore, these companies all have core businesses and don’t want to be distracted by data center construction.
Thus, the Big Four cloud computing and hyperscale tech companies are currently pursuing a strategy of continuing to build their own infrastructure while primarily relying on outsourced vendors. My other post, “Cryptocurrency Miners Successfully Transformed into Data Centers,” discusses their primary approach.
Hardware is the only profitable area for AI
Only a few software vendors have proven profitable from AI (please see my post of “Some AI software vendors are starting to make money“), but nearly all hardware vendors heavily involved in AI data centers stand to profit—including semiconductor chip design, semiconductor materials, semiconductor equipment, servers, network communications, memory, storage media, data center infrastructure providers, and even utilities like electricity and water.
Here are some typical examples:
- Seagate Technology has surged 156% this year, becoming the best-performing stock in the S&P 500. Rival Western Digital ranks third with a 137% gain, and Micron Technology is fifth with a 93% gain. For detail, please see my post of “Seagate comeback thanks to AI, No 1 S&P 500 perform stocks YTD“
- Power producer Vistra Corp. has risen 53% this year, following a 258% surge in 2024. Please see my post of “Why did US largest electricity Vistra, a turned around company, share return higher than Nvidia?“
- Chipmaker Broadcom has a market capitalization of $1.6 trillion.
- Oracle, driven by demand for cloud computing services, has become the tenth-largest company in the S&P 500 by market capitalization. Following its earnings report on September 9, Oracle’s stock surged 36% in a single day, reaching a valuation not seen since the dot-com bubble.
The Demand for AI Is Unlimited
According to a CNBC report on August 20, OpenAI CFO Sarah Friar stated that the company achieved $1 billion in monthly revenue for the first time in July, setting a new record. Annual revenue is projected to reach $12.7 billion, but demand for computing power continues to surge. OpenAI said it achieved $1 billion in monthly revenue for the first time in July. “The demand for GPUs and computing power right now is staggering,” she told CNBC’s “Squawk Box” on Wednesday. “The biggest problem we face is a persistent shortage of computing power, which is why we launched the Stargate project and why we’re expanding our buildout.“
Friar said the growing demand for computing power requires more diversified risk diversification and increased supply, citing collaborations with Oracle and Coreweave, but also mentioned Microsoft’s continued involvement.
OpenAI CEO Sam Altman said in July that the company will have more than one million GPUs by the end of the year. By comparison, Elon Musk’s xAI announced that it was using a supercluster called Colossus, which boasts over 200,000 GPUs, to train Grok4.
2029 spending exceeded expectations by $80 billion, reaching $115 billion, due to OpenAI’s increasing cloud rental and computing costs.
Risks are rising
Stargate is not progressing smoothly
The $500 billion Stargate project has also been delayed due to protracted negotiations with other partners and a lack of finalized site selection. When the Stargate project was announced at the White House in early 2025 by Trump, Oracle, OpenAI, and SoftBank, they pledged an immediate $100 billion investment, aiming to reach $500 billion over the next four years.
However, as of the end of July 2025, no data center plans had been finalized, and the project’s progress was delayed. OpenAI and SoftBank disagreed over the terms of their partnership, with OpenAI even signing contracts with other data center providers. Also noteworthy is SoftBank’s original plan to establish a joint venture with OpenAI in Japan to provide AI services to local corporate clients, but progress on this project has significantly lagged.
However, Stargate’s core goal is to promote American re-industrialization while simultaneously suppressing the rise of Chinese AI, creating direct competition between China and the US – a seemingly insurmountable challenge. With the emergence of more high-quality Chinese AI companies, Stargate’s long-term competitive advantage and chances of success remain uncertain. Stargate’s success will be tested by multiple factors, including funding, technology, and politics.
AI investment and capex are staggering
In 2025 alone, the largest US tech companies, including Meta, will invest over $155 billion in AI development. According to Statista, the current AI market is valued at approximately $244.2 billion.
Amortization and Depreciation
Because all of this has occurred within the last three years, accounting rules mean that amortization and depreciation have not yet begun. Future amortization and depreciation will inevitably erode these companies’ profits.
AI Irrational Exuberance
MIT’s study
Is there a risk of repeating the economic crises caused by overinvestment in the internet in the 1850s and 2000s? An MIT study in August found that despite over $40 billion invested, 95% of AI pilot projects have failed to achieve a return on investment. Just before the MIT report was released, OpenAI CEO Sam Altman sounded the alarm about an AI bubble, expressing concern about the inflated valuations of some AI startups and excessive investor enthusiasm.
Zuckerberg insists FOMO
What about the AI bubble? Zuckerberg: Missing out is the biggest disaster, not wasting hundreds of billions of dollars. “For companies like Meta, the risk may be not being aggressive enough, not being too aggressive.” “There are compelling reasons why AI may be an outlier,” Zuckerberg responded on the Access podcast. “If the capabilities of the models continue to improve year after year and demand continues to grow, then perhaps there won’t be a collapse.” It’s not just Sam Altman warning about an AI bubble. Now Mark Zuckerberg says a collapse is definitely a possibility.
OpenAI’s opinion
“When bubbles occur, smart people get overly excited about subtle nuances of truth,” Altman told The Verge, adding that artificial intelligence is a reality: it’s transformative and real, but often surrounded by irrational enthusiasm. Altman also warned that “a frenzied pursuit of anything labeled ‘AI'” could lead to overvaluations and create risks for many companies.
The consequences of these bubbles can be costly. During the dot-com bubble, investors, driven by hype and enthusiasm for emerging internet companies, had unrealistic expectations for tech startups and piled into them. When the results fell short of expectations, the total market value of stocks in the dot-com bubble evaporated by over $5 trillion.
OpenAI will invest trillions of dollars in infrastructure. Altman: The AI craze involves irrational investment but also contains real technology. The AI field is experiencing unprecedented enthusiasm, and the frenzied development is sparking heated debate about whether it’s a bubble. Bret Taylor, chairman of OpenAI, ChatGPT’s parent company, recently stated that while we’re currently in an AI bubble, this won’t prevent AI from ultimately creating significant economic value.
In a recent interview with The Verge, Taylor echoed OpenAI CEO Altman’s earlier observations, stating, “We’re in an AI bubble, and some people are going to lose a lot of money.”
He also warned that, like any disruptive technological wave, AI’s development will inevitably produce both huge winners and significant losers. While AI will transform the economic landscape and create enormous value, it can also coexist with market bubbles.
Taylor directly contrasted the current AI craze with the dot-com bubble. He mentioned that many companies collapsed after the bursting of the Internet bubble, but in the long run, “those people in 1999 were correct to some extent in their judgment of the future of the Internet. Today, companies born in that era, such as Amazon and Google, have become among the companies with the highest market capitalization in the world, which shows that the vision under the bubble can be realized.”
Related articles
- “AI data center shows no signs of cooling down in the near future“
- “Tech giants’ AI strategy are shifting“
- “AI jeopardizes cash flow and profits“
- “AI job openings tripled in past five years“
- “Seagate comeback thanks to AI, No 1 S&P 500 perform stocks YTD“
- “Some AI software vendors are starting to make money“
- “How does Applied Digital make money?“
- “Why did US largest electricity Vistra, a turned around company, share return higher than Nvidia?“
- “A must-read for Nvidia investors＂The Thinking Machine＂“
- “How GPU farms CoreWeave make money?“
- “How does Comfort Systems, unattractive business but amazing stock performance, make money?“
- “Data center, a rapidly growing semiconductor field“
- “Artificial intelligence benefits industries“
- “How Much Does Generative AI Cost?“
- “How Vertiv, share price return 2.5 times of Nvidia, make money?“
- “Why did US largest electricity Vistra, a turned around comapny, share return higher than Nvidia?“
- “Top five lucrative artificial lucrative intelligence listed companies“
- “Supermicro, a repeat offender of scandals, valuation is not justified and unsustanable, no worth for long-term holding“
- “The artificial intelligence bubble in the capital market is forming“
- “Artificial intelligence investment trap“
- “Major artificial intelligence companies in US stocks market“
- “OpenAI, the Generative Artificial Intelligence rising star and ChatGPT“
- “Top five lucrative artificial lucrative intelligence listed companies“
- “Arm and Qualcomm’s IP licensing lawsuit set to be settled“
- “Arm relisted successfully and it’s ushering in its best days“
- “How does the ubiquitous Arm make money?“
- “Why nVidia failed to acquire ARM?“
- “How much potential does ARM’s WOA (Windows on ARM) has?“
Disclaimer
- The content of this site is the author’s personal opinions and is for reference only. I am not responsible for the correctness, opinions, and immediacy of the content and information of the article. Readers must make their own judgments.
- I shall not be liable for any damages or other legal liabilities for the direct or indirect losses caused by the readers’ direct or indirect reliance on and reference to the information on this site, or all the responsibilities arising therefrom, as a result of any investment behavior.
