InvestAnswers
InvestAnswersSep 20
Tech

Age of AGI 💀 TSLA x SPCX LEAPs, Anthropic Short, Chinese Humanoid Stocks

39 min video5 key momentsWatch original
⚡
TL;DR

CAPE ratio is obsolete in the age of AGI — traditional valuation metrics don't apply when AI creates exponential returns, so don't let Warren Buffett's playbook scare you out of compounders.

Key Insights

1

LEAP timing beats buy-and-hold — Tesla 140 calls bought at stock bottom near 106 are now 14x returns with 89 days left — bought at 16 bucks, now worth 224, versus just 2.1x on the stock itself.

2

Merger odds 72% by 2028 — Kelshi prediction markets show 72% odds of Tesla-SpaceX merger by 2028, 37% by May 2027 — no deal announced yet but if stock deal happens, LEAP positions get adjusted by ratio.

3

Open source wins revenue later — Open-source AI takes 79% of tokens but closed models still capture 21% of revenue — dynamic will flip as enterprises build sovereign AI on open weights to avoid OpenAI IP theft.

4

Closed models unsustainable long-term — OpenAI losing 300 billion over next three years while Anthropic is overpriced; houses of cards unless they sustain growth — but infrastructure buildout on open models continues regardless.

5

Buffett lost the tech race — Berkshire Hathaway underperformed S&P 500 for 18 straight years because Buffett feared technology — CAPE metric breaks down entirely in exponential growth era.

6

Compute supremacy beats models — He or she with most compute wins AI race, not model weights on thumb drives — SpaceX controls 30-40% of Vera Rubin supply and will make 100 billion ARR from data center hosting.

Want this for every new video InvestAnswers posts? Brevyd summarizes each upload automatically, the morning it drops.

Deep Dive

Why CAPE is Dead in the AGI Era

Anderson W challenges James with the CAPE ratio sitting at 41 versus historical median of 16, asking if a violent correction looms. James responds firmly: CAPE doesn't work when exponentiality enters the equation. The ratio was built for railroad and industrial economies, not for a world where AI creates infinite ROI machines. His example: a Tesla Cybercab costs 15 grand, generates 50k yearly — that's infinite return. He cites Warren Buffett as a cautionary tale. Buffett was a legend until year 2000, then spent 25 years terrified of technology. His Berkshire Hathaway underperformed the S&P 500 for 18 consecutive years. The QQQ delivered 31x over 20 years despite two crashes. James concludes that CAPE-based caution will trap you out of the actual wealth creation happening in AI compounders.

LEAP Strategy: Buy Beaten Down, Sell at Peaks

James walks through his Tesla 140 calls purchased near 106 stock price two years ago for 16 bucks each. They're now worth 224 with 89 days until expiration, representing a 14x return versus 2.1x for owning stock outright. He hammers the core rule: never buy out-of-the-money LEAPs unless stock is at an extreme bargain. Example from the community: someone looking at SpaceX 500 calls when stock trades at 140 is lunacy. James shows Marvel following identical playbook — stock gave 2.1x but the 20 dollar LEAP is now 167 (8.35x), and the synthetic long had a 33.4x return. The trigger for buying is ruthless: wait for big selloffs, watch IDSS crystal ball chart for buy flags, then accumulate on dips. He bought when Marvel was crushed, added on second dip, rode it up. The key: LEAPs only work if underlying asset will produce massive returns — Tesla, Marvel, AMD, Broadcom all fit that bill.

Anthropic Short and Open vs Closed AI Moat Debate

Mr. Burns asks if James would short Anthropic at IPO, and whether regulatory capture creating a duopoly could flip the thesis. James is clear: he doesn't trust Dario Amodei, sees signs of a sociopath, and generally refuses to buy into companies run by people he distrusts. However, he'll make an exception if risk-reward screams opportunity — like he did with Meta at 90 bucks. On the bigger question of open versus closed winning, James lays out both sides. Closed models (Apple, walled garden) historically dominate and capture pricing power. Regulatory capture could accelerate that. But his thesis is that model weights fit on a thumb drive — no durable moat there. Open-source can't be banned; it spreads like wildfire, like drugs in prison. Enterprises won't build on OpenAI or Anthropic platforms because those companies steal IP and learnings. Instead, big enterprises build sovereign AI on open models running their own infrastructure — the IA13 pick-and-shovel play. Token share favors open 79% to 21%, but closed still wins revenue for now. Eventually margins can't exist in free market, and the shift accelerates.

Chinese Humanoid Robotics: Sana, Invance, XPeng Plays

Jesse Dave raises the USA versus China AI race and whether humanoids are worth playing in China. James agrees it's a two-horse race, nobody else contending. China has two aces: can produce at scale faster than anyone except Elon, and has way more electricity (critical for compute). James owns small positions in Baba and BU for full transparency, but notes Chinese stocks carry delist risk, lower disclosure standards, and much lower valuations than US peers. On humanoid robotics specifically, he maps the food chain: Sana makes actuators, Invance has servos and drivers, HSAI does eyes with LiDAR, XPeng (XPv) has cash engine plus robotics division, UB has ripping revenue and is Hong Kong-based, Horizon Robotics makes chips for robots. Sana is back at 2022 levels and could easily double. Invance is back at 2021 levels. But all carry massive risk — many are lossmaking. James final take: skip the individual Chinese names and just buy Tesla if you want humanoid exposure. It's simpler and far less risky than chasing Sana or Invance.

SpaceX Compute Dominance and Path to 100B ARR

A live caller asks if Grok makes financial sense and whether SpaceX compute hosting can break out as separate profit line. James pivots to fundamentals. Grok owns most coherent compute globally, is putting data centers in space, has global comms. More importantly, they can replicate frontier model weights by end of year — watch for Grok 4.7, 4.8, 5.0 releases. SpaceX itself will generate 100 billion ARR from compute hosting. James cites OpenAI's CFO at a May event saying compute is unavailable through end of 2027 except through Elon's operation. Nobody can spin data centers faster, build energy supply, own mega packs, and control 30-40% of Vera Rubin supply. That's total dominance. He or she with most compute wins the AI game, not model weights on thumb drives. This frames SpaceX IPO and potential Tesla merger as vehicles to own that compute supremacy story.

Takeaways

  • ✓Buy LEAPs only when stock is beaten down 30-50% and has path to 3-5x returns — never chase them at market tops or on out-of-the-money strikes.
  • ✓Ignore CAPE ratio valuation signals in exponential growth environments; Buffett's 18-year underperformance proves old playbooks break in AGI era.
  • ✓If you hold 30% cash waiting for a crash, you're leaving exponential gains on the table — deploy into compounders while predicting markets say 72% merger odds by 2028.
  • ✓Skip individual Chinese humanoid names and buy Tesla instead if you want robotics exposure — Sana, Invance carry delist and disclosure risks that don't justify the upside.

Key moments

8:00Tesla LEAP Returns Crush Stock Gains

“I bought these when the stock was beaten down near 106. And now after these couple of years, they're about to expire. So if I just bought the stock, I'd make a 2.1x of my money today from where the price is today. This LEAP cost 16 bucks. And that's a 14x return.”

24:00CAPE Ratio Broken in AGI Era

“CAPE does not work in the age of AGI. We are going I use the term exponentiality. This isn't a railroad or the invention of the internet. This is way bigger.”

26:00Buffett Lost to Technology Fear

“Berkshire Hathaway has failed to beat the S&P 500 over the last 18 years. You're better off just buying an ETF, an index ETF. The legend ran out of juice, you know, quarter of a century ago at the turn of the century.”

33:00Open Source Dominates Token Share

“Open is about 78% 79% versus closed about 21%. This is moving the margin from the model layer to the infra and apps.”

38:00SpaceX Compute is Only Game in Town

“If anybody tells you they can get their hands on compute through the end of 2027, they are lying. The only place you can actually get it is Elon Musk's SpaceX and they're selling it and they're making a ton of money on it.”

You just read one. Brevyd does this for every upload.

Follow InvestAnswers and every new video comes back as a summary like this, in your morning briefing. No watching required.