ClaimsSatya Nadella

“we built an entire rack which is liquid cooled for Maia.”

Satya Nadella, Microsoft CEO 15 November 2023 CNBC Transcript: Nadella on The Exchange (CNBC) archived copy

Factual AI compute, hardware and data

Microsoft built an entire liquid-cooled rack for its Maia AI chip.

Resolved true The record shows it happened or was so.

Factual: Statements about the present or the past, checked against the record: did they hold up?

Said for an organisation they lead or work for; it counts on both records.

What would count as true or false

Judge as of 2023-11-15. True: Microsoft disclosed by about that date that it had designed or built a custom server rack, liquid-cooled (for example with a dedicated liquid-cooling 'sidekick' unit), specifically to house its Maia AI accelerator. Mostly true: a Maia-specific rack existed, but a minor detail differs, for example the cooling was a hybrid design or the rack was built with partners. Mostly false: Maia racks existed, but were not liquid-cooled or not built specifically for Maia. False: the evidence shows no such rack had been built.

Set before the research began.

Evidence (4)

  1. E1 Microsoft is finally making custom chips — and they’re all about AI The Verge · 15 November 2023

    “Microsoft built a unique rack to house Maia server boards in, complete with a “sidekick” liquid chiller that works like a radiator you’d find in your car or a fancy gaming PC to cool the surface of the Maia chips.”
  2. E2 Microsoft is finally making custom chips — and they’re all about AI The Verge · 15 November 2023

    ““Maia is the first complete liquid cooled server processor built by Microsoft,” reveals Borkar.”
  3. E3 Microsoft looks to free itself from GPU shackles by designing custom AI chips | TechCrunch TechCrunch · 15 November 2023

    “Microsoft says that it had to build from scratch the data center server racks that house Maia 100 chips, with the goal of accommodating both the chips and the necessary power and networking cables.”
  4. E4 Microsoft’s Maia AI, Azure Cobalt chips to rev up efficiency, performance Network World · 15 November 2023

    “To implement liquid cooling, Microsoft has developed a “sidekick” that sits next to the Maia 100 chips’ rack.”

Reasoning

The Verge and TechCrunch reported on 2023-11-15 that Microsoft had built a custom rack for Maia (E1, E3), and Network World described a liquid-cooling sidekick next to the Maia 100 rack (E4). The Verge describes a unique rack for Maia server boards with a "sidekick" liquid chiller (E1). TechCrunch reports that Microsoft built the racks for Maia 100 from scratch (E3). Network World describes a liquid-cooling sidekick unit that sits next to the Maia 100 rack (E4). This meets the true criterion, which names a sidekick unit as an example of the liquid cooling it requires.

Cite this claim

Satya Nadella, “Microsoft built an entire liquid-cooled rack for its Maia AI chip.” (said 15 November 2023). Verdict: resolved true. Reliability API, claim c_c5f7beb43778, method 0.1.0+28da144b, dataset release v1.1.2. https://epistacklabs.com/claims/c_c5f7beb43778

This claim is in dataset release v1.1.2, with the same verdict.

Reply to this claim

For Satya Nadella, or someone speaking for them: a correction or counter-evidence. Replies are read before anything is shown; an approved reply shows your name, role, date, statement and link. Your email address is never shown and is deleted 30 days after the decision (privacy).

More factual claims by Satya Nadella

  1. Mixed Factual 28 January 2026
    “same thing in inferencing as well.”

    AI inferencing, like AI training, requires a lot of compute and a lot of storage located close to that compute.

  2. Mostly true Factual 28 January 2026
    “we’ve been at this in a variety of different forms for a long, long time in terms of building our own silicon.”

    Microsoft has been building its own silicon in various forms for a long time.

  3. Resolved true Factual July 2025
    “Through software optimization alone, we are delivering 90% more tokens for the same GPU compared to a year ago.”

    Through software optimization alone, Microsoft delivers 90% more GPT-4o-family tokens per GPU than a year ago.

  4. Mostly true Factual 15 November 2023
    “right as we speak, we have some of the Maia stuff powering GitHub copilot for example.”

    As of November 2023, some Maia chips are powering GitHub Copilot.

  5. Mixed Factual 4 June 2026
    “I was able to use a model that is using 500B, I was able to use a 5B, and have it really deliver the same outcome, why would I not use that?”

    In a Land O'Lakes agent example shown at Build, a 5B-parameter model delivered the same outcome as a 500B-parameter model.

All of them

Verdicts are made automatically; how verdicts are made.