Nvidia just told the world that the party isn't over yet. Despite the sudden absence of the man in the leather jacket, the momentum behind the world's most critical compute infrastructure seems to have enough inertia to carry it through the next fiscal year. For builders and founders, this isn't just about stock prices; it's about the roadmap of the chips we use to train our models.
The Shadow of the Founder
Losing a visionary CEO like Jensen Huang under a cloud of controversy is usually a death sentence for a tech giant's short-term guidance. Usually, leadership vacuums lead to conservative estimates and cautious outlooks. Nvidia is doing the opposite. They are leaning into the demand curve, banking on the fact that the industry's hunger for H100s and the upcoming Blackwell architecture is larger than any one personality.
For those of us building in AI, this tells us two things. First, the supply chain is likely stabilized enough that they feel confident in their delivery timelines. Second, the institutional demand from the hyperscalers—the Microsofts and AWSs of the world—is not slowing down just because the C-suite changed. The infrastructure layer of the AI stack is now a commodity so vital that it transcends individual leadership.
The Blackwell Transition
The real story isn't the revenue numbers; it is the transition from Hopper to Blackwell. Every founder I talk to is trying to figure out if they should buy compute now or wait for the next generation. Nvidia’s bullish revenue projections suggest that the Blackwell ramp-up is going better than the rumor mill suggested. If they expect revenue to keep soaring, it means they have solved the thermal and yield issues that were supposedly plaguing the new chips.
If you are a startup founder, this means the cost of compute might not drop as fast as you hoped. If demand is still outstripping supply to the point where Nvidia can lose its founder and still project record growth, your margins are still at the mercy of the hardware layer. We are still in the 'build at any cost' phase of the AI cycle.
Why Skepticism is Still Required
I am naturally skeptical when a company under fire promises a golden future. It is a classic move to keep the share price from cratering. However, in Nvidia's case, the orders are already on the books. They aren't selling software subscriptions that can be canceled tomorrow; they are selling physical silicon that takes months to manufacture and ship. These revenue projections are backed by purchase orders from companies that have already committed billions to their data center build-outs.
The risk for builders is not that Nvidia fails, but that we are over-building for a demand that hasn't materialized at the application layer yet. Nvidia is making money selling the shovels. Whether the gold miners—the LLM wrappers and agentic workflows we are all building—actually find gold is still an open question.
The Builder's Perspective
We need to stop treating Nvidia as a proxy for the entire AI industry. They represent the energy cost of the movement. If their revenue is soaring, our overhead is staying high. As founders, we should be looking for ways to optimize our inference costs. If the hardware is staying expensive, the software must become more efficient.
- Don't assume compute will get cheaper in the next 12 months.
- Watch the Blackwell rollout closely; if delivery slips, the secondary market for H100s will stay inflated.
- Focus on small language models (SLMs) that don't require the top-tier hardware that Nvidia is currently bottlenecking.
The departure of a founder is a cultural shift, but in the world of semiconductors, the lead times are too long for it to matter this quarter. The machines are already in motion. Nvidia is betting that the momentum of the AI revolution is now greater than the reputation of its leadership.
What Happens Next
We are going to see a lot of noise about the 'new Nvidia' and how it handles the post-Jensen era. Ignore most of it. Watch the data center revenue segment. If that number keeps climbing, it means the big players are still terrified of being left behind. As long as that fear exists, Nvidia wins.
For the rest of us, the goal remains the same: build something people actually want to use so that when the hardware cycle eventually cools down, we have a real business left standing. The infrastructure is being built at a record pace; now we just have to make sure something useful lives inside it.
Read the original at TechCrunch AI →