He’s going to cast his own blades and vanes Eager to dig into this tomorrow. But saying +10GW with capacity for +20GW I wonder what total capacity is for B&V today and how long it took to stand up

Another +10 GW of compute for SpaceX/xAI in 2027 is physically and technically possible in definition. Delivering it on that timeline, alongside everything else Musk is taking on, seems ambitious to say the least. They reported 1.7 GW of nameplate compute at September-end, up from 1.4 GW in June. Adding 10 GW next year means averaging 2.5 GW per quarter, over 8x last quarter’s additions. That’s more than 3x their expected ~3 GW year-end fleet added in one year. Using the high-end SemiAnalysis’s ~50 GW global forecast for 2027 additions as a rough benchmark, one player would account for ~20% of the buildout. The power definitions need to match (e.g., unclear if Elon is using IT or facility load; if IT add another 20% GW for the cooling/electrical/overhead which amplifies the problem), but that’s a huge share of an already exceptionally bullish industry forecast. If 10 GW means IT load, an illustrative ~3 kW per GPU, including allocated system overhead, implies ~3.3 million GPUs (which begs of the Q of can you get that many chips). At an assumed $30–50 billion per IT GW, it would also mean $300–500 billion of investment (which begs the Q of can you finance them in this environment). They’ve demonstrated exceptional construction speed. Industrial conversions and behind-the-meter generation give them a credible way around long grid timelines. There may be much more equipment secured and construction underway than is publicly visible. But turbines still leave gas delivery, land, permits, transformers, cooling, construction crews and commissioning. Buying acreage in Texas, Mississippi and Tennessee seems achievable. Getting enough sites approved, equipped and operating on that schedule is much harder. They’re competing with hyperscalers and other AI developers for the same constrained chips, equipment and skilled labor. The political environment is getting harder too. Power bills, water, noise and emissions are feeding bipartisan opposition, moratoriums and litigation. SpaceX’s existing Southaven power project is already defending an environmental lawsuit. Financing also faces greater scrutiny. CoreWeave’s 2031 bond yield touched almost 16% last week as investors reassessed its debt risk. For projects requiring this much capital upfront, delays extend the gap between spending and revenue. All of that while bringing more manufacturing in-house and scaling Starship. A Tesla merger, if it happens, would add another major undertaking. I’ve learned to avoid calling Elon’s targets impossible. I’m still skeptical of this deadline without seeing how much of the next 10 GW already has sites, permits, equipment allocations and funded construction behind it. Late here, eager to dig further into the project pipeline tomorrow.

Another +10 GW of compute for SpaceX/xAI in 2027 is physically and technically possible in definition. Delivering it on that timeline, alongside everything else Musk is taking on, seems ambitious to say the least. They reported 1.7 GW of nameplate compute at September-end, up from 1.4 GW in June. Adding 10 GW next year means averaging 2.5 GW per quarter, over 8x last quarter’s additions. That’s more than 3x their expected ~3 GW year-end fleet added in one year. Using the high-end SemiAnalysis’s ~50 GW global forecast for 2027 additions as a rough benchmark, one player would account for ~20% of the buildout. The power definitions need to match (e.g., unclear if Elon is using IT or facility load; if IT add another 20% GW for the cooling/electrical/overhead which amplifies the problem), but that’s a huge share of an already exceptionally bullish industry forecast. If 10 GW means IT load, an illustrative ~3 kW per GPU, including allocated system overhead, implies ~3.3 million GPUs (which begs of the Q of can you get that many chips). At an assumed $30–50 billion per IT GW, it would also mean $300–500 billion of investment (which begs the Q of can you finance them in this environment). They’ve demonstrated exceptional construction speed. Industrial conversions and behind-the-meter generation give them a credible way around long grid timelines. There may be much more equipment secured and construction underway than is publicly visible. But turbines still leave gas delivery, land, permits, transformers, cooling, construction crews and commissioning. Buying acreage in Texas, Mississippi and Tennessee seems achievable. Getting enough sites approved, equipped and operating on that schedule is much harder. They’re competing with hyperscalers and other AI developers for the same constrained chips, equipment and skilled labor. The political environment is getting harder too. Power bills, water, noise and emissions are feeding bipartisan opposition, moratoriums and litigation. SpaceX’s existing Southaven power project is already defending an environmental lawsuit. Financing also faces greater scrutiny. CoreWeave’s 2031 bond yield touched almost 16% last week as investors reassessed its debt risk. For projects requiring this much capital upfront, delays extend the gap between spending and revenue. All of that while bringing more manufacturing in-house and scaling Starship. A Tesla merger, if it happens, would add another major undertaking. I’ve learned to avoid calling Elon’s targets impossible. I’m still skeptical of this deadline without seeing how much of the next 10 GW already has sites, permits, equipment allocations and funded construction behind it. Late here, eager to dig further into the project pipeline tomorrow.

I'm re-sharing this ahead of tonight's season opener, before LeBron begins YEAR 24 at AGE 41. I expect his averages to decline in a supporting role. That should do nothing to diminish what he's already produced. The fact we're still discussing his role on a contender is objectively insane. Jordan is the greatest scorer I've watched, and I give him the rate/peak edge. LeBron is the best all-around player I've watched. He could carry the scoring, run the offense, rebound and defend across positions, then change his game as the roster and the league changed. For one game, one series or a fully healthy championship run, it's fair to lean Jordan. But, there are very credible LeBron arguments for each. Over an entire career, I don't see a toss-up. LeBron gave his teams far more years of elite scoring and creation. That's why he's my GOAT. Full article, interactive charts and sources: shanumathew.com/writing/jord… cc: @bronhistory @LeBronTeam @BronMuse @WITNESSKJ @kinglbj_23

LeBron played 123 more playoff games than Jordan. Every additional run gives opponents another chance to expose a weakness and gives you another chance to play badly with everyone watching. Across 302 playoff games, he averaged 28.2 points, 8.9 rebounds and 7.2 assists. His lowest scoring average across 19 postseasons was 23.2. He leads the all-time playoff totals in points, games, minutes and steals. When his team faced elimination, he averaged 32.8 points, 10.4 rebounds and 7.3 assists across 30 games. In Game 7s: 34.9 points, 9.9 rebounds, 5.6 assists and a 6-2 record. He also has five playoff buzzer-beating game-winners.