A one-gigawatt AI factory started at $20-30 billion, is now $50-60 billion, and will soon be $80-100 billion per gigawatt.
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?
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
The claim has three parts: about $20–30B per GW at the start, about $50–60B per GW now, and $80–100B per GW soon. Judge each against industry estimates of total build cost per gigawatt of AI data-centre capacity as of 2026-06-02, not the cost per GW at a later date. The later 'soon' trajectory counts only as a stated expectation that is reasonably supported. True: all three are supported, with figures within roughly 20% of the stated ranges and the forward figure consistent with credible projections. Mostly true: the trend and two figures hold, with one figure modestly off. Mixed: roughly half the parts hold, for example the historical and current figures are right but the forward projection is unsupported. Mostly false: there is only a rising trend, and the cited figures are substantially off. False: the per-GW cost estimates contradict the claimed levels and trend.
Set before the research began.
Evidence (4)
-
E1 Introducing the Frontier Data Centers Hub Epoch AI · 4 November 2025
“We estimate a typical capital cost of $44B per gigawatt of server power, with $30B going to IT hardware and $14B to other costs. But the cost per gigawatt of total facility power is lower, typically closer to $30B, because overheads decrease the number of servers that can be supported.”
-
E2 AI’s Growth Spurred Infrastructure Investment Worldwide DeepLearning.AI (The Batch) · 26 December 2025
“Each gigawatt of data-center capacity will cost roughly $50 billion to build.”
-
E3 Bigger than the railroads: The staggering money behind AI’s data center buildout - The Boston Globe The Boston Globe · 26 October 2026
“Van Nieuwerburgh, in an October working paper published by the National Bureau of Economic Research, calculates that a facility drawing 1 gigawatt of power — enough for about 850,000 homes — costs about $41 billion.”
-
E4 Dylan Patel & Stacy Rasgon: Semiconductor Industry Outlook panel Fireside Alpha (Substack) · 24 October 2026
“Patel says a gigawatt cost “roughly $40 billion” to build in 2024 and “Now it’s north of 60,” and that it is “not clear who’s going to eat all the cost.””
Reasoning
The current figure of $50–60B per GW has some support. E2 (Dec 2025) puts the cost at roughly $50B per gigawatt, and E1 estimates $44B per GW of server power, which is within about 20% of the range. The starting figure of $20–30B gets partial support: E1 puts the cost per GW of total facility power at about $30B, but E4 has Patel saying a gigawatt cost about $40B in 2024, which is above the claimed range. None of the evidence gives a credible projection of $80–100B per GW, so the forward part is unsupported. Roughly half the claim holds, which makes it mixed.
Cite this claim
Jensen Huang, “A one-gigawatt AI factory started at $20-30 billion, is now $50-60 billion, and will soon be $80-100 billion per gigawatt.” (said 2 June 2026). Verdict: mixed. Reliability API, claim c_abfd6cb1e193, method 0.1.0+03d23c0e, dataset release v1.1.2. https://epistacklabs.com/claims/c_abfd6cb1e193
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More factual claims by Jensen Huang
-
Mixed
“the reason why NVIDIA’s growth is accelerating is because people are starting to realize that we’re the safest architecture to invest in.”
-
Mostly true
“Trillions of dollars in investment.”
Current investment in AI is on the order of trillions of dollars.
-
Mostly true
“Today, it’s dollars per million tokens.”
AI intelligence is now monetized and priced in dollars per million tokens.
-
Mostly true
“Now CoreWeave is worth $50, $60, $70 billion dollars and growing incredibly fast.”
CoreWeave is now worth $50, $60 or $70 billion and is growing very fast.
-
Mixed
“Today, we announce four new partners joining the NVIDIA RoboTaxi Ready platform: BYD, Hyundai, Nissan, and Geely, with a combined annual production capacity of 18 million vehicles.”
Verdicts are made automatically; how verdicts are made.