The past 48 hours have been nothing short of extraordinary—stock prices crashing, claims of a bursting bubble, and plenty of end-of-days rhetoric. But before we start preparing for the apocalypse, let’s take a step back and ask: Who is this really affecting?
Unlike some market crashes that hit every industry, this downturn targets a specific group of companies—namely those that produce large language models (LLMs) and those that supply them with critical technology (like GPUs) and utilities (particularly power). That might be alarming if you’re a technology supplier or an LLM producer. But what about everyone else?
Cheaper Supply, Bigger Opportunity
Most organizations don’t build their own LLMs; they use them as a commodity resource, much like we consume electricity, water, or any other utility. And history has plenty of lessons about what happens when a critical supply becomes dramatically cheaper.
1. Steel (Mid-19th Century)
The Bessemer Process slashed the cost of producing steel, enabling mass production of high-quality steel from pig iron. Railroads, construction, and manufacturing all boomed as they gained access to more affordable, high-grade steel. Carnegie Steel (later U.S. Steel) and railroad companies capitalized on this, laying the foundation for rapid industrial expansion.
2. Aluminum (Late 19th Century)
With the invention of the Hall-Héroult Process came a significant reduction in the cost of aluminum production, leading to aluminum’s widespread adoption. Aerospace, packaging, and construction industries flourished thanks to lighter, more durable materials available at lower prices.
3. Ammonia (Early 20th Century)
Synthetic ammonia—crucial for nitrogen-based fertilizers—became cheaper to produce using the Haber-Bosch Process, revolutionizing agriculture. Farmers worldwide benefited from more affordable fertilizers, leading to higher crop yields, lower food prices, and the ability to feed a rapidly expanding global population.
In each case, a significant drop in the cost of a key input led to an explosion of growth in related industries. If generative AI suddenly becomes far more affordable—perhaps by 95%, as some are suggesting—we can expect a similar pattern of innovation and expansion.
As CTO of an organization building deep and broad AI capabilities, this represents a huge opportunity to me. And it might not mean that we use DeepSeek, but their simply existing will change the pricing dynamic across the whole industry.
Overcoming the “It’s in China” Concern
Some skeptics argue that enterprises won’t adopt certain open-source LLMs because they’re hosted in China or because they contain politically sensitive filters. In reality, by the time this blog is published, numerous providers in the US and Europe will have likely spun up hosting solutions for these models—complete with guarantees that no data leaves their region.
Sure, the model might be selective about discussing certain political or historical topics, but is that a real hindrance if you’re primarily using it to build data products and applications? When you can reduce your AI costs by up to 95%, it’s hard to argue that these constraints outweigh the benefits.
The Market Agrees
I’m not the only one who sees it this way. While GPU and AI-focused stocks took a hit, Snowflake’s stock actually rose over 6% in the same period. This divergence suggests that investors recognize the potential upside for companies (like Snowflake or DataOps.live) that help businesses use LLMs rather than build them.
Some believe this marks the start of an “AI crash,” but history hints at the opposite. We may look back and see this moment as a turning point—where the economics of generative AI shifted, unlocking far broader applications.
If you’d like to dive deeper, there’s an excellent (though paywalled) article on the subject that you can access if you have one free monthly read. Here’s to an exciting future, where generative AI becomes more accessible, affordable, and transformative than ever before!
