AI AI, Cap’n: Shifting headwinds to the promised land of tech innovation

In any industry, the simple bits tend to be the most stable. For every ice cream, there’s the company that makes the sticks; for every book publisher, there’s the company selling the paper. By the same logic, in the midst of the AI boom, Nvidia – which manufactures chips that make the technology possible – has been the beating heart, making up about 8% of the S&P 500.

Nvidia’s Q2 2025 earnings were strong, up 6% on Q1 and 56% year-on-year. However, within two days of the Q2 earnings report, the company’s stock price was down 4%, according to NPR. And it wasn’t alone.

Semiconductor company Marvell Technology fell 18.9%. The Philadelphia Semiconductor index, which tracks 30 of the world’s biggest manufacturers, was dragged down to its worst performance since mid-April, according to the Financial Times.

When the shovels stop selling, what does that say about the gold rush?

To be fair, it isn’t that the “shovels” aren’t selling – it’s that they’re not making as much of a profit as investors would hope, with revenue up but still failing to meet forecasts (per the FT).

The models are a different story, however. ChatGPT, the most-used generator by far on a consumer level according to MIDiA survey data, pulls in billions of annual revenue, and is still loss-making. Its competitors are in the same boat. The costs of running them are high, from the chips themselves, to the energy demands to power their data centres, and the water needed to cool them. This will continue to be difficult to offset as consumers are now accustomed to accessing it for free, and, as of yet, there are no routes to ad revenue.

If gold cost more to mine than you could make by selling it, the shovel industry wouldn’t do so great, either.

The road to El Dorado is paved with profit incentives

Promises of job growth and an economic renaissance powered by AI have been critical to prospective investment and development. Yet it is starting to seem like this prospective investment and development is the economic renaissance, with little coming out the other side. A study by MIT has found that 95% of generative AI pilots at companies are failing (per Fortune). CNBC reports that Meta has announced a hold on investment for now (although it has attributed this to organisational planning, rather than the market shift).

The promise of AI – in entertainment especially, but not exclusively – is that it can make higher-quality output cheaper to create. Yet it is on its own a very expensive product, with costs unlikely to go down unless the prices of electricity, water, shipping, and manufacturing all decrease significantly (in which case, we’d have much bigger things to think about). The last few years of AI integration have been largely experimental, with the new technology shoehorned in wherever it might fit. Yet with such a high cost that will eventually need to be rebalanced, this is not sustainable. These second-quarter earnings and subsequent stock adjustments hint that we are finally at the turning point where promises are no longer enough: shareholders need to see results.

Fools gold cannot build kingdoms

The creative industries have been faced with deeply existential concerns as a result of AI. The SAG-AFTRA strikes were an early pushback from creators; recent lawsuits by Disney, Universal, and Warner Bros. Discovery show even major studios are growing cautious (not to mention the ongoing Anthropic case). Social media mainly stands to benefit as the de-facto home of AI creation: for them, more content should (in theory) mean more value, and more business, reflected especially by Meta’s initial embrace of it.

Yet for music, TV, and social alike, AI has exacerbated a serious preexisting problem: the value of content itself. Streaming drove the price of accessing all the music and movies in the world down to a mere $10 per month. Social flooded friends’ feeds with creator hot takes, disrupting genuine connections. Value and fandom hinge on the underlying IP around which content is created, yet competition and cut-through were already huge challenges – now exponentially increased by the proliferation of AI content. Last June, Europol estimated that 90% of content on the internet would be all or partly AI generated by 2026; one might argue that we’re already there.

AI’s disillusionment moment means that, on the business side, its use must become more strategic. However, because it remains accessible to the public, content proliferation will continue, especially on social platforms. We were already in an environment of attention saturation and hyper-competition, with consequential needs for gatekeeping and better curation. These problems need addressing, no matter what happens with AI itself.

The reality is that AI’s best uses are not going to be that exciting. For example, it can function as a smart algorithm or search, helping with curation and personalisation. It can potentially help with licensing and attribution in a similar way, simplifying the process.

Yet as the digital space becomes ever more cluttered with content and ever more personalised of an experience, moments of connection between people will become harder than ever to find. This may be the biggest impact on entertainment. Fandom is a shared experience, and cultural movements require collective momentum. Addressing this conundrum primarily means turning to new, curation- and community-focused spaces and experiences, both on and offline. Despite all the hype, AI is merely one tool in the kit.