Author of *The Bitter Struggle* on US-China AI competition: chip export policy is bipartisan and China is not waking up yet

Teddy Collins on The Bitter Struggle: chip policy is bipartisan, China isn't panicking yet

Teddy Collins co-authored The Bitter Struggle while working on AI policy inside the Biden White House, and the book's core argument is that democratic nations face three compounding problems: maintaining a lead in AI capabilities, closing the gap between government and industry on how to use the technology, and ensuring it doesn't erode democratic values as it matures.

The publishing timeline is a caveat worth flagging upfront. Collins and his co-author locked the text near the start of 2025, which means it predates several significant AI developments. Anyone reading it today will notice the gaps.

“In 2021, when we ended the administration, AI was not a thing that most policymakers were really paying attention to. The first struggle is for democratic nations to establish and maintain a leading capability. The second is for governments to understand and make use of this technology effectively. The third is to make sure that the technology doesn't become deeply unsafe and doesn't take us away from democratic values.”

Chip export controls

Collins takes a broadly laissez-faire starting position on export restrictions, but argues frontier AI chips sit closer to the nuclear-weapons end of the spectrum than the Coca-Cola end. His time in government reinforced rather than softened that view, though he acknowledges drawing the exact line, specifically which nanometer node thresholds matter, consumed enormous amounts of cabinet-level attention in ways that felt almost absurd in the moment.

On the politics, the narrative that chip hawkishness is a Republican position is outdated. The initial chip sanctions began under Trump's first term, were layered on further during Biden, and most of the core restrictions have survived into Trump's second term. Collins says Congress has shown genuine bipartisan support for the policy when it's framed carefully rather than as nativist reaction. The messier picture is at the executive level, where the views of Commerce, Treasury, and the president himself appear to diverge in ways that are hard to track in real time.

China's posture

Collins is skeptical that China has had a genuine strategic awakening on AI. Chinese policymakers and industry leaders are talking more about AGI and ASI than they were a couple of years ago, but the modal discourse in China still treats AI as a significant but broadly normal technology. The CEO of DeepSeek is an exception, having framed the company around AGI from founding, but that framing is not representative of the broader conversation.

The deeper problem is opacity. China's internal debates don't surface publicly, so reading its strategic intent requires Cold War-style Kremlinology, watching who gets invited to present to the Politburo Standing Committee rather than reading public statements. Collins cites work by researcher Jeff Ding as a corrective to easy survey-based comparisons between Chinese and American AI enthusiasm, arguing those comparisons are usually apples to oranges and that the gap in public attitudes would likely be smaller than reported if measured under identical conditions.

On the theory that China favors open-source AI because it commoditizes capabilities where the US holds an advantage, Collins is skeptical. Someone still has to pay the compute bill regardless of whether a model is open source, and China's unemployment pressures limit how far state infrastructure spending can absorb labor displacement.

Nvidia's Shanghai workforce

Asked directly about Nvidia's roughly 4,000 employees in Shanghai doing R&D, software development, and sales, Collins declines to speculate on outcomes. His honest position is that the entanglement between US and Chinese companies, through supply chains, employees, and customers, is far messier than the sharp bilateral competition framing suggests, and predicting where each relationship lands as new lines get drawn is genuinely unknowable right now.

Inherent

Collins is building Inherent, a London-based AI lab focused on scientific discovery and R&D collaboration. The core pitch is that science shouldn't be treated as a task to automate and hand to AI, but as a human-machine teaming problem. Inherent targets domains where rewards are nonverifiable and underspecified, areas where human taste and judgment still outperform current systems, and aims to build agents and platforms that keep humans meaningfully in the loop.