Why are AI industry giants suddenly seeking 'smaller' data centers?

Imagine you have hired the world’s most brilliant chef. This chef creates masterpieces but requires a massive kitchen and thousands of specialized knives. What would happen if this chef grew into a team of 10 or 100? Rather than mindlessly building one gargantuan kitchen, creating efficient, smaller kitchens in neighborhoods everywhere would be much faster and cheaper.

Recently, AI industry giants OpenAI and Anthropic have been grappling with this exact dilemma. Until now, they have staked everything on securing massive, hyperscale data centers—gigantic facilities housing AI servers—to train their models. However, news has emerged that they are shifting their strategy to actively seek “small-to-midsize data centers.” Source 1, Source 2

Why does this matter?

While it might sound like a simple change in construction scale, this shift signals that the AI industry has reached a crucial inflection point. In the past, the core competition was “who can build the bigger model,” but now, “how efficiently can we operate the model and deliver it closer to the user” has become the priority.

For the AI services we use to become faster and more stable, infrastructure must be flexible. Relying solely on hyperscale facilities could leave entire services vulnerable if a problem occurs in a specific region. Conversely, distributing smaller data centers reduces risk and allows for immediate responses from locations physically closer to users. In short, AI technology is beginning the practical preparation to move beyond research labs and embed itself deeply into our daily lives.

Simply put: Data centers are AI’s ‘food warehouses’

To use an analogy, for high-performance AI models like Transformers (the brain structure of AI that understands sentence context), data centers are like warehouses for storing food.

Until now, AI companies focused on building dozens of massive warehouses. Source 4, Source 9 But if a warehouse is simply too large, it becomes difficult to manage, and if even one freezer in a specific warehouse breaks down, a massive amount of food is at risk of spoiling all at once.

Their current search for smaller warehouses is to deliver the “food” more quickly to where it is needed. Placing warehouses that can be tapped immediately in various locations is far more economical and safer than relying on a single, distant, giant warehouse. Distributed data centers allow AI models to be much more agile when performing complex coding or intricate tasks. Source 7

Current situation: Giants expanding their territory

Both OpenAI and Anthropic are growing at breakneck speeds in terms of revenue. Anthropic is nearing an annual recurring revenue (ARR) of $9 billion, while OpenAI has recorded staggering growth, approaching nearly $40 billion. Source 9, Source 11

For companies moving such massive capital, infrastructure is their lifeblood. Anthropic is making aggressive moves, such as hiring expert negotiators to secure computing resources in the European region. Source 3 OpenAI, meanwhile, is actively securing infrastructure by collaborating with SoftBank and Nvidia to build massive data centers in Ohio. Source 9 They are now building “infrastructure portfolios” that encompass both hyperscale and mid-sized facilities.

What happens next?

Experts evaluate these moves as a shrewd strategy to maximize infrastructure efficiency. This is because the business of building and operating data centers is structured to generate returns that far exceed the initial investment. Source 5

Moving forward, the two companies are expected to go beyond the competition of simply increasing capacity and stake their success on how efficiently and how broadly they can secure their “AI food warehouses.” For users like us, this is good news, as it increases the likelihood that AI services will be provided faster and without interruptions. However, we must also watch closely to ensure that the monopoly on infrastructure by these giants does not become a “high barrier” that prevents new competitors from entering the market. Source 6

MindTickleBytes’ AI Reporter Perspective

Ultimately, the core of the AI race is shifting from “who creates the smarter model” to “who manages more energy efficiently.” This pivot to smaller data centers is evidence that AI is moving beyond a mere experimental lab toy and firmly establishing itself as a massive infrastructure for our society.

References

1. Anthropic, OpenAI hunt for smaller AI data center deals, sources say - https://www.cnbc.com/2026/09/18/anthropic-openai-small-ai-data-center-deals.html
2. Anthropic and OpenAI reportedly seek smaller data center deals… - https://digg.com/tech/7c20becf-4731-452b-8525-0f81863432f1
3. Anthropic looks to hire six-figure role for negotiating data center… - https://benzatine.com/news-room/anthropic-seeks-key-negotiator-for-european-data-center-expansion-amid-ai-boom
4. Anthropic’s Little Brother - The Atlantic - https://www.theatlantic.com/technology/2026/04/openai-imitating-anthropic/686975/
5. Dylan Patel – Anthropic & OpenAI will have most of the… - https://www.dwarkesh.com/p/dylan-patel-3
6. AI इंसानों को मार देगा - खतरा असली या मोनोपली का खेल? OpenAI… - YouTube - https://www.youtube.com/watch?v=JG__sioj_ZY
7. Research \ Anthropic - https://www.anthropic.com/research

8. IPO Fever Heats Up for OpenAI and Anthropic

   The Motley Fool - https://www.fool.com/investing/2026/08/22/ipo-fever-heats-up-for-openai-and-anthropic/

9. Anthropic Leaks ARR of $65B Fueling IPO Chatter… - https://www.linkedin.com/pulse/anthropic-leaks-arr-65b-fueling-ipo-chatter-openais-lag-buchanan-c2zee
10. Home \ Anthropic - https://www.anthropic.com/
11. The revenue race between Anthropic and OpenAI is getting more heated - https://www.voronoiapp.com/business/The-revenue-race-between-Anthropic-and-OpenAI-is-getting-more-heated-6929