BG2• Jul 31, 2025• 1:04:20Interview

China Open-Source, Compute Arms Race, Reordering Global Trade | BG2 w/ Bill Gurley and Brad Gerstner

From BG2

Brad Gerstner(Host)•Bill Gurley(Host)•Sonny Madra(Guest)

Executive Summary

China is rapidly advancing in open-source AI, with models like Qwen challenging top US proprietary models and creating a new competitive front in the global AI race.
A massive 'cost is no object' compute arms race is underway, with companies like OpenAI, X.AI, and Google planning unprecedented investments in data centers and GPUs, driven by exploding demand for AI inference.
Venture capital is pouring into AI labs at historic valuations, fueling a market-share-focused strategy with negative gross margins and raising questions about long-term business model sustainability.
The US is pursuing an aggressive new trade policy, securing favorable deals with Europe and Japan, and signaling a potential for a major, market-moving trade reorientation with China.

12 quotes

Concerns Raised

China's rapid, compounding progress in open-source AI could allow it to surpass US proprietary models.
The 'cost is no object' spending and negative gross margins of major AI labs are unsustainable long-term.
Excessive US regulation in energy, AI, and semiconductors has been an 'unforced error' that threatens American leadership.

Opportunities Identified

Explosive, 200x growth in AI inference demand creates a massive market for compute infrastructure and services.
A potential major US-China trade deal could reorient the global economy and create a significant market tailwind.
The rise of powerful open-source models creates opportunities for new companies to build sanctioned, enterprise-ready AI solutions.

Key Themes

The AI Arms Race: US vs. China

China has emerged as a dominant force in open-source AI, with multiple high-quality models that are rapidly improving by compounding on each other's work. This progress, exemplified by models like Qwen reaching GPT-4o performance, poses a significant challenge to US leadership, which has been hampered by self-imposed regulations.

This dynamic shifts the competitive landscape from a US-dominated proprietary model race to a dual front, where open-source accessibility and rapid iteration could erode the moats of incumbent AI leaders.

The Unprecedented AI Compute Buildout

The demand for AI inference is growing exponentially, as seen in Google's token processing volume skyrocketing 200x in about a year. This is fueling a massive arms race for compute power, with OpenAI and X.AI planning to spend hundreds of billions on multi-gigawatt data centers and millions of GPUs.

This 'sport of kings' creates enormous opportunities for the AI supply chain (NVIDIA, energy, data center infrastructure) but also raises the barrier to entry for new foundational model companies, concentrating power among a few hyper-scalers.

AI Business Models & Venture Capital Mania

AI labs like Anthropic are raising billions at staggering valuations (e.g., $170B), driven by insatiable demand. However, this capital is being used to pursue market share at all costs, leading to negative gross margins and pricing below cost.

The current market dynamic is unsustainable. It suggests an eventual market correction or a shift where defensibility comes from user lock-in and high switching costs rather than pure model performance, especially with cheap open-source alternatives available.

US Geopolitical & Trade Realignment

The US administration is executing a high-risk, high-reward trade strategy, securing deals with Europe and Japan that involve unilateral tariffs and massive investment commitments to the US. This aggressive posture is now being turned towards China, with expectations of a major deal that could rebalance the global trade relationship.

This proactive and often confrontational trade policy creates both volatility and opportunity. A successful grand bargain with China could de-risk global markets and unlock growth, while a failure could trigger damaging trade wars.

The Open-Source vs. Proprietary AI Divide

The AI landscape is increasingly defined by the tension between proprietary models and the burgeoning open-source ecosystem, largely dominated by Chinese players. Chinese open-source models offer near-proprietary performance at a fraction of the cost, creating intense pricing pressure.

This commoditizing force threatens the business models of proprietary AI labs. It forces all players to develop an open-source strategy and creates opportunities for new companies to build enterprise-grade solutions on top of powerful, low-cost models.

Get started free

Topics

AI Open Source AI China AI US-China relations Geopolitics Trade Policy Tariffs Venture Capital AI Compute NVIDIA OpenAI Anthropic Grok Reasoning Models Data Centers Inference Qwen Alibaba Crypto Regulation Stablecoins

Processed Apr 9, 2026 yt-dlp + mlx-whisper + Gemini