More AI Hardware Competition Is a Tailwind for $IREN Yesterday, OpenAI released the first measured performance results for Jalapeño, its first custom inference chip and the first generation of a multi-generation silicon platform being developed alongside $AVGO. The early numbers are impressive, with OpenAI claiming significant advantages over Nvidia’s GB200/GB300 systems in throughput per watt and latency across several LLM inference workloads. My good friend @jiahanjimliu published an excellent article breaking down the nuances behind those results and, importantly, what they do and don’t mean for $NVDA. His argument is that while Jalapeño can outperform Nvidia GPUs in more narrowly defined inference workloads, it remains a far more specialized piece of hardware. Nvidia GPUs offer substantially greater versatility across workloads, making Jalapeño much less of an immediate threat to Nvidia than the headline benchmark numbers might suggest. I largely agree. But from an $IREN shareholder’s POV, I think there is another important angle here: More competition in AI hardware is ultimately a tailwind for cloud providers like $IREN. While Nvidia remains the undisputed king of AI hardware, it’s becoming increasingly clear that its largest customers, particularly hyperscalers and frontier labs, are trying to verticalize more of the stack by developing their own silicon. That dynamic helps explain why $NVDA has such a strong strategic interest in supporting open-source and open-weight models, as well as the broader neocloud ecosystem. A world with significant enterprise demand for open-weight models makes it much harder for a handful of frontier labs to monopolize token output on proprietary models running on proprietary silicon. Likewise, a deep and competitive neocloud ecosystem built around Nvidia hardware reduces Nvidia’s dependence on hyperscalers as a customer cohort. In other words, $NVDA appears very aware of where its biggest long-term threats are coming from and is actively building an ecosystem that hedges against them. On the other side of that equation are Nvidia’s customers, who ultimately stand to benefit from greater competition in AI hardware. We are still in the early innings, and those benefits aren’t particularly material yet. But over time, greater hardware competition should give cloud providers more optionality, improve their negotiating leverage, and put downward pressure on hardware prices. To be clear, I don’t think Nvidia’s margins are under meaningful threat today. They could remain extraordinarily strong for years. But from the perspective of a cloud provider, having your most important hardware supplier operate as something close to a monopoly is never ideal. One of the most tangible examples, particularly for neoclouds like $IREN, $CRWV, and $NBIS is memory. As HBM and other component costs increased, Nvidia was able to pass that inflation through to customers via higher system pricing. Eventually, that dynamic should work in reverse as memory supply expands and costs come down. The question is how much of those savings Nvidia chooses to pass through. In a quasi-monopoly, there is no guarantee that lower input costs translate proportionally into lower prices for customers. With credible hardware competition, that equation changes. If Nvidia refuses to pass through enough of the savings, competitors can, making their hardware increasingly attractive on a relative cost basis. Nvidia may well pass those savings through regardless. But competition makes the incentive much stronger. For cloud providers, that ultimately means more suppliers, more negotiating leverage, lower hardware costs and potentially better returns on deployed capital. All that said, this industry remains in its infancy, and $NVDA is still the undisputed king of AI compute. I don’t expect that to change anytime soon. But the strategic battle lines are becoming increasingly clear. I’ll be paying close attention to $NVDA's earnings call today, and I’m obviously even more interested in what $IREN has to say tomorrow. Cheers!