Meta's whole GPU fleet is the roughly 350,000 GPUs the interviewer cited as planned by end of year.
Mixed Partly true and partly false in roughly equal measure.
Factual: Statements about the present or the past, checked against the record: did they hold up?
The evidence is the person's or their organisation's own records.
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
Judged as of 2024-04-18. The proposition is that the roughly 350,000 GPU figure the interviewer cited is Meta's whole GPU fleet planned for the end of 2024, and not, for example, a training-only cluster. True: Meta's statements show that the ~350,000 figure (Meta's announced target of about 350,000 NVIDIA H100s by end-2024) covers its entire fleet across all uses, including training, inference and ranking/recommendation. Mostly true: it is the whole fleet of one GPU type, and other compute is minor or described in a way consistent with 'whole fleet'. Mixed: the figure is a fleet-wide count of H100s, but Meta also described substantially more total compute (for example about 600,000 H100-equivalents), so 'whole fleet' is only partly accurate. Mostly false: the 350,000 is a meaningful subset of a clearly larger fleet. False: the 350,000 refers only to a training cluster or a single project.
Set before the research began.
Evidence (3)
-
E1 META Q4 2023 Earnings Call Transcript Meta Platforms, Inc. · 1 February 2024 Own record
“I recently shared that by the end of this year we'll have about 350k H100s and including other GPUs that’ll be around 600k H100 equivalents of compute.”
-
E2 Building Meta’s GenAI Infrastructure Meta Engineering (engineering.fb.com) · 12 March 2024 Own record
“By the end of 2024, we’re aiming to continue to grow our infrastructure build-out that will include 350,000 NVIDIA H100 GPUs as part of a portfolio that will feature compute power equivalent to nearly 600,000 H100s.”
-
E3 Mark Zuckerberg indicates Meta is spending billions on Nvidia AI chips CNBC · 18 January 2024
“Mark Zuckerberg said on Thursday that by the end of 2024, the company's computing infrastructure will include 350,000 H100 graphics cards.”
Reasoning
In Meta's Q4 2023 earnings call (E1), the company said that by the end of 2024 it would have about 350,000 H100s, and that "including other GPUs" its compute would be around 600,000 H100 equivalents. Meta's engineering blog (E2) also calls the 350,000 H100s part of a portfolio with compute equal to nearly 600,000 H100s. So the 350,000 figure is the fleet-wide H100 count, not the whole GPU fleet, and the total compute is substantially larger. That matches the criteria's "mixed" case.
Cite this claim
Mark Zuckerberg, “Meta's whole GPU fleet is the roughly 350,000 GPUs the interviewer cited as planned by end of year.” (said 18 April 2024). Verdict: mixed. Reliability API, claim c_639f756195b3, method 0.1.0+3f84d14d, dataset release v1.1.2. https://epistacklabs.com/claims/c_639f756195b3
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More factual claims by Mark Zuckerberg
-
Resolved true
“So we’re building out many, many gigawatts of compute.”
-
Mostly true
“Most of that is due to higher component costs, particularly memory pricing.”
-
Mostly true
“we are rolling out more than one gigawatt of our own custom silicon that we are developing with Broadcom as well as a significant amount of AMD chips to complement the new NVIDIA systems that we are…”
-
Resolved true
“I recently announced MetaCompute with the belief that being the most efficient at how we engineer, invest, and partner to build our infrastructure will become a strategic advantage.”
Zuckerberg recently announced MetaCompute, Meta's infrastructure initiative.
-
Resolved true
“The people who are joining us will have access to unparalleled compute as we build out several multi-GW clusters.”
Meta is building out several multi-gigawatt compute clusters.
Verdicts are made automatically; how verdicts are made.