We in the "West" have a complicated relationship with intellectual property. We celebrate it when it protects our inventions. We condemn it as theft when others use the same techniques we employ ourselves. Nowhere is this tension more visible than in two industries that will define this century: renewable energy and artificial intelligence.

China has now played a pivotal role in the world twice through what Western companies call intellectual property theft. First, by manufacturing solar panels at a scale that has made clean energy affordable globally. Second, by developing AI models that are driving the cost of intelligence toward zero.

The outrage from Silicon Valley is loud. But the pattern is unmistakable. When Western companies want to move fast, they operate under "ask forgiveness, not permission." When Chinese companies do the same, it's suddenly theft.

Act One: How "Stealing" Solar Technology Is Saving The Climate

Germany led the charge on solar power in the early 2000s. Companies like SolarWorld flourished on government subsidies. They invested heavily in research, developed cutting-edge technology, and sold panels at premium prices. It was a textbook case of innovation through patient capital and state support.

America took a different path. Betting on hydraulic fracturing, the United States poured resources into shale gas. The gamble paid off handsomely. According to the International Energy Agency, Chinese industrial policies focusing on solar PV as a strategic sector and on growing domestic demand enabled economies of scale and supported continuous innovation throughout the supply chain, contributing to a cost decline of more than 80%.

China, lacking domestic oil and gas reserves, saw an opportunity. In 2010, Europe and the United States held significant market shares in solar panel production. However, aggressive government policies, lower labour costs, and massive infrastructure investments in China fundamentally reshaped the industry. The methods were straightforward: acquire the technology through licensing, partnerships, and various transfer mechanisms, then scale production to levels no other country could match.

The result? China dominates global production with approximately 80-85% market share. More remarkably, based on manufacturing capacity under construction, China's share of global polysilicon, ingot and wafer production will soon reach almost 95%.

Was this intellectual property acquisition controversial? Undoubtedly. Did it accelerate climate solutions? Quite possibly.

Enter China: No Oil, No Choice, No Patience

According to Carbon Brief, China's CO2 emissions fell by 1% in the final quarter of 2025, likely securing a decline of 0.3% for the full year as a whole. This extends a "flat or falling" trend in China's CO2 emissions that began in March 2024 and has now lasted for nearly two years. Solar power output increased by 43% year-on-year, wind by 14% and nuclear 8%, helping push down coal generation by 1.9%.

This is the world's largest emitter responsible for nearly a third of global CO2 reaching a critical turning point years ahead of its 2030 target. The cheap solar panels that made this possible didn't emerge from decades of patient Western capital. They came from China's industrial strategy of acquiring and scaling technology relentlessly.

Meanwhile, China can pump out a full terawatt worth of solar panels each year. That capacity isn't staying in China. It's flooding global markets, making renewable energy affordable in countries that could never have accessed it at Western prices.

The AI parallel: history repeating at Silicon Valley's expense

Now the pattern is repeating with artificial intelligence.

According to Anthropic's official statement, the company is accusing three Chinese AI companies of setting up more than 24,000 fraudulent accounts with its Claude AI model to improve their own models. The labs, DeepSeek, Moonshot AI, and MiniMax, allegedly generated more than 16 million exchanges with Claude through those accounts using a technique called "distillation." Anthropic said the labs "targeted Claude's most differentiated capabilities: agentic reasoning, tool use, and coding."

Distillation is simple in principle. Query a sophisticated AI model millions of times. Save the responses. Use that output to train your own model. The student learns from the teacher without paying tuition.

Distillation is a common training method that AI labs use on their own models to create smaller, cheaper versions, but competitors can use it to essentially copy the capabilities of other labs.

Anthropic frames this as theft. The timing is notable. While not identical to distillation, Anthropic was recently accused of copyright violations by thousands of authors, allegedly downloading books in bulk from shadow libraries to train its AI models, rather than buying copies and scanning them itself. In a historic move, Anthropic settled that lawsuit for $1.5 billion in September 2025.

The company built its models in part using pirated books from shadow library sites, according to the settlement. It also bought and scanned additional volumes. It relied on datasets that a court found violated copyright. The settlement resolved the plaintiffs' claims.

Now that same company accuses Chinese labs of theft.

The double standard is striking. When Western AI companies take without permission from writers, photographers, and artists, it's innovation. Fair use. The cost of progress. When Chinese companies query Western AI models at scale, it's suddenly industrial espionage requiring "a coordinated response among industry players, policymakers, and the global AI community."

Act Two: How "Stealing" AI Models is Killing the CAPEX Bubble 

The consequences of Chinese competition are already visible in the economics of AI. For example, Chinese AI lab DeepSeek scored 96.0% on AIME 2025 while charging $0.28 per million input tokens (cache miss), roughly one-tenth the cost of comparable Western models.

Performance is comparable. Costs are a fraction. The model is open-source under an MIT licence.

This poses significant challenges for the AI business model that Silicon Valley has been building. And the market knows it.

OpenAI is revising its AI infrastructure ambitions. According to CNBC, the ChatGPT maker is now telling investors it plans to spend roughly $600 billion on compute through 2030, a 57% reduction from the $1.4 trillion in infrastructure commitments CEO Sam Altman touted just months ago.

Now reality is asserting itself. The recalibration comes as the company faces mounting questions about whether its revenue growth can ever justify the scale of its spending ambitions, and signals a potential inflection point in the AI infrastructure cycle.

I'm Not Defending Chinese Governance

Let me be clear: I'm not a fan of the Chinese government's approach to governance for a variety of reasons. But we need to be honest about the hypocrisy at play here.

American AI companies built their models on pirated books, stolen articles, and copyrighted content they never paid for. Now they're upset that Chinese companies are doing to them exactly what they did to authors and publishers—except the Chinese version is arguably less harmful because it doesn't even require destroying the original.

The race to zero: when intelligence becomes a commodity

Chinese AI models demonstrate something Silicon Valley may find uncomfortable: frontier AI performance is increasingly becoming a commodity.

The economics are challenging. If you can distil a model by querying it millions of times, the value of closed-source models may trend toward zero. The capital expenditure required to train frontier models becomes harder to justify when competitors can replicate performance at a fraction of the cost through distillation.

This is similar to what happened with solar panels. Western companies invested in R&D, built the initial technology, and charged premium prices. Chinese manufacturers acquired the knowledge, scaled ruthlessly, and drove prices down significantly. Many companies that pioneered the technology struggled or exited the market. But the world got cheap solar panels. And now we get cheap, commoditised AI, again saving us from unnecessary electricity consumption and mindless CAPEX spending on datacentres.