Dinner with.. @AlphaOnTheMenu

COFFEE WITH SK HYNIX ☕ $SKHY SK hynix remembered my order. Then asked whether I’d like to reserve the next three years of coffees. Me: “Business must be good?” SK hynix: “Q2 revenue grew 257%. HBM4 mass shipments started. Long-term agreements with around ten customers.” Me: “Most cafés are happy with a loyalty card.” SK hynix: “Our customers prefer guaranteed supply.” Me: “Speaking of customers—what’s brewing with Intel?” SK hynix: “Reuters reports talks about using part of Intel’s Ohio facility, or a joint venture that could include major cloud companies.” Me: “So the people ordering the coffee might help pay for the machine?” SK hynix: “Possibly. Nothing’s agreed.” Me: “And making it in America?” SK hynix: “More expensive.” A customer-backed venture could spread the funding burden and tie new capacity to longer-term buying commitments. That would give SK hynix better visibility than building a factory first and hoping demand arrives later. US production could also strengthen its relationships with customers seeking a local supply. But those benefits depend on the terms. Who pays for construction? Who absorbs overruns? Are customers committing to purchases, or simply expressing interest? Higher US costs, potential Korean regulatory review and the memory cycle still matter. More capacity can eventually weaken the pricing that makes today’s expansion attractive. My bullish case starts with the existing AI-memory business and HBM4 progress. Intel adds an interesting potential route to growth—but I’m not counting an unsigned deal as future earnings. Imagined dialogue, not company quotes. $INTC

DINNER WITH GOOGLE 🍽️ $GOOGL Google had already picked the restaurant and knew what I wanted. I hadn’t opened the menu. Me: “How’s business?” Google: “Q2 revenue grew 24%. Cloud grew 82%. Search & other grew 17%.” Me: “Thoughts on Dario’s tweet about slowing down AI?” Google: “We’ve announced at least €13bn for Finnish AI infrastructure in 2027–28. Three new data centres. A 22-year nuclear-power agreement.” Me: “The construction team hasn’t received the slowdown memo, then.” Slower frontier development doesn’t necessarily mean less infrastructure spending. Me: “One more thing. Have you achieved recursive self-improvement? AI getting better at improving itself?” Google: Takes a long sip of wine. I’m bullish without needing the RSI rumours to be true. Search is still growing, Cloud is capturing AI demand, and Finland adds computing capacity backed by long-term power supply. Google has several ways to earn from AI. The risk is the price of that expansion: capital spending exceeded operating cash flow this quarter. For me, the opportunity justifies that near-term pressure—but the investment needs to deliver stronger cash generation over time.

DINNER WITH META 🍽️ $META Meta picked the restaurant before I’d finished searching for one. Me: “Business looks busy.” Meta: “Q2 revenue rose 28% to $60.80bn.” Me: “So plenty left for dessert?” Meta: “Operating cash flow was $31.86bn. Capital spending took $31.08bn.” Me: “And what’s the next course?” Meta: “Muse. A personal AI agent—with a free tier.” Suddenly, the spending has something new to show for it. I’m bullish on Meta. Muse’s free offering looks promising: lowering the barrier to trying an AI agent could help turn curiosity into everyday use. The advertising business helps fund that ambition. The next test is whether users stick around—and whether adoption eventually produces enough income to justify the spending. Imagined dialogue, not company quotes. No trade represented.