Meta is building the Hyperion cluster, which will be able to scale up to 5GW over several years.
Unresolvable The public record could not settle it.
Promise: Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
The claim has two parts: (a) Meta was building the Hyperion data-centre cluster, and (b) the cluster can scale up to 5GW 'over several years', taken here as by the end of 2030. True: construction of Hyperion was under way, and by 2030 its capacity reached, or was firmly built out toward, about 5GW (roughly 4.5GW or more). Mostly true: Hyperion was built and reached a large fraction of 5GW (roughly 4GW or more) by 2030, or reached 5GW slightly later while the plans stayed unchanged. Mixed: Hyperion was built and is operating, but its capacity or official plans were cut well below 5GW (roughly 2 to 4GW). Mostly false: Hyperion was built only at a small scale (under about 2GW), or the 5GW scale is reached only well after 2030. False: Hyperion was cancelled or never built.
Set before the research began.
Evidence (4)
-
E1 Meta Hyperion: Full Guide to the Louisiana AI Data Center Inside Deep Tech · 28 October 2026
“Meta expects 2 GW by 2030 and the full 5 GW by around 2032, according to a company spokesperson.”
-
E2 Meta Hyperion | AI data centers Epoch AI
“Meta Hyperion is an AI data center under construction with AI chips owned and used by Meta. In Q1 2028, we project it to host an estimated 4,143k H100-equivalents of AI compute, supported by 1,632 MW of IT power at a capital cost of $61.8 B.”
-
E3 Deepening our investment in Richland Parish, Louisiana - Meta Data Centers Meta · 13 July 2026 Own record
“We’re expanding our AI-optimized data center in Richland Parish, Louisiana to 5 GW in compute capacity — the largest in Meta’s fleet”
-
E4 Meta Hyperion: Full Guide to the Louisiana AI Data Center Inside Deep Tech · 28 October 2026
“Mark Zuckerberg first used the name in July 2025, when he said Meta was building Hyperion, "which will be able to scale up to 5GW over several years", alongside a 1 GW-class cluster called Prometheus in Ohio.”
Reasoning
Part (a) is supported. Epoch AI describes Hyperion as under construction (E2), and Meta announced it was expanding the Richland Parish site to 5 GW (E3). Part (b) depends on Hyperion's capacity by the end of 2030, and none of the evidence records that. All the figures are forecasts, not records of built capacity: a Meta spokesperson expected 2 GW by 2030 and the full 5 GW around 2032 (E1), and Epoch projects about 1.6 GW of IT power in early 2028 (E2). Those forecasts point to a shortfall against the 2030 benchmark, but they don't show what was actually built by then, so the claim can't be settled under the criteria.
Cite this claim
Mark Zuckerberg, “Meta is building the Hyperion cluster, which will be able to scale up to 5GW over several years.” (said 30 July 2025). Verdict: unresolvable. Reliability API, claim c_f5ec822f8c3f, method 0.1.0+3f84d14d, dataset release v1.1.2. https://epistacklabs.com/claims/c_f5ec822f8c3f
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More promise claims by Mark Zuckerberg
-
Delivered on time
“Watermelon is basically, that’s shipping soon.”
-
Mostly true
“But we are going to be rolling out our initial set of models and products and businesses around that over the coming months.”
Meta will roll out its initial set of new AI models, products and businesses over the coming months.
-
Delivered on time
“We're going to have more to share on all of this, especially our Reality Labs work, at Connect on September 17th”
-
Delivered on time
“we're pretty close to being able to show off the first prototype that we have of that, and I'm really excited about that.”
Meta is close to showing off its first prototype of the full AR glasses.
-
Resolved true
“We have a roadmap of new releases coming that are going to bring multimodality, more multi-linguality, and bigger context windows as well.”
Verdicts are made automatically; how verdicts are made.