Serenity ties a $40B SpaceX raise back to his InP chokepoint thesis as substrate shortages surface
Serenity, the stock-picker tracked here for his habit of buying the obscure upstream suppliers that feed the AI datacenter buildout, spent Sunday connecting a headline-grabbing financing event back to the materials bottleneck at the center of his portfolio.
The SpaceX signal. His starting point was a reported $40 billion capital raise tied to SpaceX (SPCX), the Elon Musk rocket and satellite company whose long-awaited market debut Serenity has framed as the defining mega-IPO of the moment. What caught his attention was not the rocket business but where the money reportedly goes: he noted the financing is earmarked largely to procure NVIDIA (NVDA) GPUs in the first half of 2027. For Serenity, that is another data point in what he half-jokingly calls the "AI infrastructure bubble" — enormous sums committed by the world's largest companies all fighting for control of future AI capacity, with little sign of a slowdown.
Why it matters to his framework. NVIDIA sits at the core of Serenity's method not as a holding — he does not own it, viewing it as a crowded mega-cap — but as a demand signal. He treats NVIDIA's roadmap and the spending flowing toward its chips as a telegraph of where the next scarcity will land. More GPUs, in his logic, mean more demand for the memory, networking and optical plumbing around them, which in turn pulls on the upstream materials he actually buys.
The chokepoint payoff: AXT. That back-mapping led him straight to AXT (AXTI), his highest-conviction core position and a maker of indium phosphide (InP) substrates — the compound-semiconductor material used to build the high-speed lasers that carry data as light inside AI datacenters. Serenity flagged commentary from Taiwanese materials supplier VPEC describing InP substrates as being in serious short supply, naming AXT among the relevant suppliers. He paired that with a Morgan Stanley view that InP substrates, along with EML and CW lasers, will be the primary bottlenecks for optical interconnects — EML lasers serving the fastest 800G/1.6T links, and CW lasers acting as the light source for co-packaged optics, where the optical engine is placed next to the switch or GPU die to save power.
The thread lines up neatly with Serenity's long-standing AXT thesis: that whoever controls the raw-material supply gains pricing power as photonics demand ramps. He has argued AXT sits across roughly 40% of the InP supply chain end-to-end — raw material, refining, crucibles and substrates — in a duopoly with Japan's Sumitomo, and that China's export controls on gallium and related inputs quietly hobble its rivals. VPEC's emphasis that future advantage belongs to whoever can secure more raw-material supply echoes exactly the argument he has used to justify holding AXT through volatility. AXT remains his self-described most legendary call.
Takeaway. There was no new trade or position change today — this was Serenity reinforcing, rather than revising, his view. A fresh wave of GPU-directed spending and independent confirmation of an InP squeeze both point in the same direction: more pressure on the upstream materials he already owns. As always, this is derived commentary on his reconstructed views, not investment advice.