Double. That's the word doing the heavy lifting in Nvidia's latest forecast, and it's worth sitting with for a second: the company that has already become the most valuable hardware supplier in modern history is telling the market it expects to sell roughly twice as many AI chips in 2027 as it will in 2026. Not a modest uptick. Not a plateau after two frantic years of shortage and scramble. A doubling. For an industry that has spent two straight years wondering aloud when the chip crunch would finally ease, the answer from the company sitting at the centre of the entire supply chain is unambiguous: not yet, and maybe not for a long while.
That kind of growth doesn't happen in a vacuum, and this week it's colliding head-on with a much messier, much less flattering story about what all these chips are actually powering. Unredacted court filings in a closely watched copyright dispute involving a major AI assistant have pulled back the curtain on something publishers have suspected for a while and executives have been reluctant to admit out loud: when readers get their answers directly from an AI system instead of clicking through to the original source, traffic to that source doesn't dip — it collapses. The filings put the drop as high as 93% in some measurements. Let that number sink in. Not a decline. Not erosion. A near-total evaporation of the click-through economy that has funded digital journalism and independent publishing for two decades.
This is the substitution effect publishers have been warning about since the first wave of AI chatbots hit the mainstream, and it has, until now, existed mostly as a theoretical anxiety — the kind of thing raised in industry panels and open letters, easy for tech companies to wave away as overblown. The unredacted filings change that calculus entirely. This isn't speculation anymore; it's discoverable evidence, sitting inside a real legal proceeding, showing that internal executives at some of the biggest names in AI were themselves alarmed by what their own data was telling them. When the people building the technology are privately worried about the numbers, the rest of the industry doesn't get to pretend the problem isn't real.
And yet, even as that legal exposure mounts, the same executives at the centre of it are fighting hard on a second front: regulation. Several of the most prominent figures in AI and big tech have pushed back publicly and privately against proposals for an industry-funded oversight body, arguing — predictably — that it would slow innovation without meaningfully improving safety. It's the same argument that's been deployed against nearly every regulatory proposal this industry has faced, and it's landing with decreasing credibility every time a new scandal, lawsuit, or leaked filing suggests the current model of self-policing isn't working.
The enforcement gaps show up in smaller, stranger ways too. A major social media platform's oversight board recently had to step in and order the removal of two AI-generated deepfake videos — a reminder that the actual mechanics of catching and removing harmful AI content are still happening case by case, platform by platform, board by board, rather than through anything resembling a coherent, industry-wide framework. Every incident like this is a small data point in a much bigger picture: the tools are scaling exponentially, and the guardrails are being built one emergency patch at a time.
Put all of this together and you get a genuinely uncomfortable snapshot of where the AI industry actually stands right now, underneath all the keynote-stage optimism. Chip demand is accelerating without any sign of a ceiling. Legal exposure over how these systems treat other people's content and livelihoods is mounting in real, documented, litigable ways. And the loudest voices in the room — the ones with the most to gain from moving fast — are still the ones arguing hardest against anyone slowing them down long enough to figure out the rules. History doesn't have a great track record with industries that grow this fast while resisting oversight this hard, and this week's revelations suggest AI is following that pattern rather than breaking it.