The US’s top AI models lead China’s by only about 6.5 months, and a set of policies the federal government could adopt today would roughly double that lead by the end of 2027, the Institute for Progress (IFP) estimates in a report published on 19 October 2026.
The report, “How to Increase America’s AI Lead Over China”, is by Tim Fist, Georgia Adamson and Konstantin Pilz. Fist described it on X as a 23k-word research report that took six months, with an interactive model of the US lead under different policy interventions.
Business Pill · SOVEREIGN AI
A one-minute explainer of sovereign AI: who controls the model, the chips and the data. It teaches the general idea only and says nothing about any company or report in this story.
The key insight: As we read it, the Institute for Progress turns the US-China AI race into a measurable gap: about 6.5 months today, with two levers it says the government already holds, remote access to chips abroad and distillation of US models.
IFP measures the gap with the Epoch Capabilities Index, which combines many AI benchmarks into a single scale. On that measure, the report estimates that top Chinese models lag the US frontier by 6.5 months, and by about 6.6 months on software engineering time horizons.
The report says both measures predict that a Chinese AI company will first release a model with Mythos-level capabilities by the end of 2026, referring to Anthropic’s Claude Mythos.
For the end of 2027, IFP estimates its policy package would gain 6.6 months of additional lead. Its chart puts China 13.1 months behind under that package, and 17.9 months behind in what it calls an extreme hypothetical, where China’s AI compute is frozen at today’s level and distillation stops.
How far China trails the US AI frontier, in months, as the Institute for Progress estimates it: 6.5 months today; 13.1 months by the end of 2027 with the policy package IFP models, which it estimates adds 6.6 months; and 17.9 months in an extreme hypothetical where China’s AI compute is frozen at today’s level and distillation stops (IFP report and chart, 19 October 2026). These are estimates from IFP’s simplified model.
What Was Posted on X
Big new IFP report from @GeorgiaCAdamson, @KonstantinPilz, and me on the US-China AI competition.
We spent the last 6 months digging into how Chinese AI companiesโ access to compute and distillation of American models translate into the current US-China capabilities gap.
The result is a 23k-word research report, including an interactive analytical model that visualizes changes in the US AI lead under different policy interventions.
Our findings:
1. The US AI lead over China is currently only around 6 months.
2. The US could roughly double this lead, to over a year, using authorities already available to the federal government.
3. On chips: Chinaโs access to AI chips will roughly double over the next year, with chips housed in data centers overseas accounting for roughly half of Chinaโs 2027 compute.
4. On distillation: the current ~16x US training compute advantage would normally yield a 13-month lead. But we estimate adversarial distillation halves this to around 6.5 months.
5. Increasing the lead would have some obvious benefits:
– Give the US broadly a greater cyberdefensive advantage against Chinese threat actors.
– Give US AI companies breathing room to spend more R&D resources on making their systems safer and more secure, without fear of being overtaken by less safety-conscious Chinese competitors.
6. Provide the US government more leverage in negotiations with China about the future of AI.
7. The two most important interventions are:
– For BIS: Create a new end-use control to restrict Chinaโs remote access to training clusters abroad โ Chinaโs largest and fastest-growing source of frontier-scale compute.
– For DOJ/FTC: Create a narrow safe harbor to allow US companies to coordinate on countering Chinese adversarial distillation.
8. Both interventions (and all the others we discuss in the report) are things the federal government could do today.
9. To estimate the effects of each policy intervention, we model the US lead as a function of its compute advantage over China and Chinaโs ability to adversarially distill the capabilities of the US frontier.
10. The model makes many simplifying assumptions, but provides a useful way to assess and rank the impact of different policy interventions at a high level.
Tim Fist, one of the report’s three authors, announced the IFP report on X on 19 October 2026 and listed its findings, including the roughly 6-month US lead and the two interventions he calls most important. We use his post for those two priorities and his list of findings.
The US-China capabilities gap is a policy choice. China relies on the US & allies: – For AI compute, via cloud access, smuggling, legal sales – For chip production, via DUVi and HBM – For AI capabilities, via distillation
All of these could be addressed today: USG could restrict cloud access, crack down on AI chip and HBM smuggling, limit SME exports, and help companies coordinate on distillation defenses.
Our report lists just some of many more tools in the belt of USG. Some take effect immediately. Others, like SME controls restrict Chinese AI progress years into the future.
— Konstantin F. Pilz (@KonstantinPilz) Oct 19 19:45
Co-author Konstantin Pilz framed the gap as a policy choice, saying China relies on the US and allies for AI compute, chip production and, via distillation, AI capabilities. We cite it as Pilz’s own framing, not as a separate finding.
Where China’s Compute Comes From
The report estimates China will have access to around 4.6 million H100-equivalents of AI compute by the end of 2026 and 8.8 million by the end of 2027, from domestic production, legally imported US chips, smuggled US chips and US chips accessed remotely in data centres outside China.
About 35% of China’s expected chip stock for the end of 2027 is already in China, IFP says. Of the remaining 65%, about half comes from remote access, which the report says can be almost entirely restricted today with new export controls from BIS.
IFP puts remote access at around 2.8 million H100-equivalents by the end of 2027, with a range of 1.97 to 3.91 million, and says these data centres are heavily concentrated in Southeast Asia, around 75% of the total.
The Distillation Effect
The report estimates that, without distillation, today’s 16x US training compute advantage would yield a 13-month lead. Adversarial distillation of US models lets Chinese companies recapture about half of that lead, by IFP’s estimates.
IFP adds that China cannot use adversarial distillation alone to gain a lead over the US, because the practice only extracts the existing capabilities of US frontier models. That, the report says, makes compute controls the more important overall intervention.
What IFP Recommends
The report lists five policies using existing executive branch authorities: a narrow antitrust safe harbor for US companies to collaborate against adversarial distillation, new export controls on remote access to training clusters abroad, stronger due diligence against chip smuggling, updated controls on chips now legal to export to China, and steps against smuggling of high-bandwidth memory.
Fist named two as the most important: a new BIS end-use control on China’s remote access to training clusters abroad, and a narrow DOJ/FTC safe harbor for US companies to coordinate against Chinese adversarial distillation.
The report argues restrictions are unlikely to reduce total US chip sales, because demand outside China is expected to outpace supply through at least the end of 2027.
How the Authors Qualify the Model
The authors say the model is high-level and makes many simplifying assumptions, and that it should be read as a way to measure the broad impact of policy interventions rather than as a highly accurate prediction of how many months each one gains.
The Structural Read
As we read it, the report singles out remote access as the most effective channel to limit: it estimates about half of the restrictable part of China’s 2027 chip stock sits in data centres outside China.
As we read it, distillation is the report’s counterweight: IFP estimates it lets Chinese companies recapture about half of the lead that the US compute advantage would otherwise give.
As we read it, the authors present the numbers as a ranking tool, not a forecast: they say the model makes many simplifying assumptions and should be read as a measure of broad impact.
Institute for Progress report, 19 October 2026
“China cannot use only adversarial distillation to gain a lead over the US, as the practice only ever allows extracting the existing capabilities of US frontier models”
Three Implications
A MEASURED GAP IFP estimates top Chinese models lag the US frontier by 6.5 months on the Epoch Capabilities Index.
A COMPUTE PIPELINE IFP estimates China will have access to around 8.8 million H100-equivalents of AI compute by the end of 2027.
A POLICY PACKAGE IFP estimates its package of existing-authority policies would gain 6.6 months of additional lead by the end of 2027.
The framework says “While attention remains fixed on GPU shortages and frontier model capabilities, the binding constraint on AI scaling has quietly shifted to High Bandwidth Memory (HBM).”
As we read it, the IFP report touches the same layer: one of its five policies targets smuggling of high-bandwidth memory into China. The framework was published in December 2025 and does not discuss the IFP report.
What Is Not Established
The months of lead are IFP’s model estimates, not measured outcomes, and the remote-access figures carry a wide range. None of the policies has been adopted. The report is IFP’s analysis; no government agency or company named in it has responded in the sources read for this piece.
Business Engineer Framework
The AI Memory Chokepoint
A Business Engineer framework on high-bandwidth memory as a binding constraint on AI scaling.
The Institute for Progress estimates that the US AI lead over China is about 6.5 months and that existing federal authorities, chiefly controls on remote access to chips abroad and a safe harbor against distillation, could roughly double it by the end of 2027.
A note on sourcing. This piece is built on the Institute for Progress report “How to Increase America’s AI Lead Over China” (19 October 2026), read in full, and on Tim Fist’s X post announcing it, including the report chart attached to that post. All months-of-lead and compute figures are IFP’s model estimates. Nothing here is a forecast of ours, and nothing here is financial or investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
