OpenAI’s latest “bizarre” mega-deal, previewing Tesla’s big announcement and more | E2189

Key Concepts: Project Stargate, Par Value, Round-Tripping, Compute Drought, Jevons Paradox, Hypercar, Downward Draft, Black Body Radiation, Special Purpose Vehicles (SPVs), Carry, MAG 7, Founder Product Fit, Founder Market Fit, Onshoring, Gross Profit Margin, Operating Profit Margin, Net Profit Margin, Generative AI, Large Language Models (LLMs), Total Addressable Market (TAM), Monthly Recurring Revenue (MRR), Customer Acquisition Cost (CAC), Lifetime Value (LTV), Average Revenue Per User (ARPU), Churn, Conversion Rate, Net Promoter Score (NPS), HIPAA Compliance, EHR Ready.

---

OpenAI's Strategic Compute Acquisitions and Market Dynamics

The discussion begins with OpenAI's aggressive infrastructure push, dubbed Project Stargate, involving significant deals with major chip manufacturers. Two weeks prior, OpenAI announced a $100 billion deal with Nvidia, which appeared to be a share trade for hardware, with Nvidia investing in OpenAI.

A new, non-traditional deal with AMD was then detailed:

• GPU Supply: AMD will supply OpenAI with up to six gigawatts worth of GPUs in several tranches, with the first delivery expected in the latter half of 2026, coinciding with the release of AMD's new 4150 GPU.
• Stock Purchase: As part of this deal, OpenAI gains the right to purchase AMD stock at 1 cent per share (par value) in tranches, eventually acquiring about 10% of AMD. This would make OpenAI the largest single shareholder in AMD.
• Funding Source: The hosts questioned the source of the $90 billion potentially spent on AMD hardware, speculating that OpenAI might be using funds from Nvidia's investment to buy AMD shares.
• Market Reaction: AMD's stock saw a significant reaction, rising by approximately 40% on the news.
• Concerns: The equity components of these deals raised concerns about round-tripping (investing in a company that then buys your products/shares), which is a "big no-no" and could attract scrutiny from regulators like the SEC, especially as OpenAI aims to go public next year.
• Non-Traditional Problem, Non-Traditional Solution: It was acknowledged that OpenAI faces a chronic "compute drought" and might need non-traditional methods to secure capacity. However, the deals are seen as circular, risky, and potentially over-leveraged if AI service demand doesn't meet the pre-purchased capacity.
• Jevons Paradox: The conversation touched upon the Jevons Paradox, an economic principle where increased efficiency of a resource (like AI compute) can lead to increased total consumption rather than decreased. This was likened to the dot-com era's fiber optic buildout, suggesting a potential oversupply of compute capacity if demand levels off.
• AI Spending: It was noted that 80% of stock market gains this year reportedly came from the AI trade, indicating massive AI spending.
• OpenAI Usage: During its Dev Day, OpenAI announced ChatGPT now has 800 million weekly active users, implying a continued need for significant compute capacity.

Tesla's Roadster 2.0 Teaser and Hypercar Technology

A new teaser video from Tesla was discussed, featuring a spinning fan blade and the number "107." This was speculated to be related to the upcoming Roadster 2.0, a "vanity project" that is "super late on Elon time" but expected to be "extraordinary."

• Downward Draft Fans: The key speculation was that the sound and visual indicated a downward draft fan system for extreme high-speed performance. This technology is crucial for maintaining grip at high speeds, preventing cars from "flying" due to wind getting underneath.
• McMurtry Speirling Example: The McMurtry Speirling, a hypercar (costing over $1 million, capable of 200-250+ mph), was cited as an example. This car can "literally drive upside down" due to fans underneath that generate 2,000 kilograms of on-demand downward force from zero mph. The Speirling holds the fastest lap record on the Top Gear test track, beating an F1 car by five seconds.
• Other Guesses: Other possibilities for the teaser included a quiet leaf blower (a past Elon Musk joke) or an advanced HVAC system, given Tesla's focus on optimizing car components for battery efficiency.
• Top Speed Records: The discussion also touched on electric cars like the BYD YangWang 09 Extreme hitting 308 mph (500 km/h) and the potential for hybrid cars (like the Corvette ZR1 X) to combine electric motors for quick acceleration with internal combustion engines for sustained top speeds.

The Vision of Space Data Centers

Jeff Bezos's vision for gigawatt data centers in space was highlighted:

• Advantages: Bezos predicts these will be built within 10-20 years, leveraging 24/7 solar power without clouds, rain, or weather, and the natural cold of space for cooling. He believes this will eventually "beat the cost of terrestrial data centers."
• StarCloud (formerly Lumen Orbit): This startup, previously featured on "This Week in Startups" (episode 2073), aims to build space-based data centers.
  • Origin: The idea stemmed from their initial work on space-based power, realizing 95% of beamed power would be lost, so it made sense to put the compute where the power is.
  • Technology: They plan to use large modular 4 km x 4 km solar arrays and proprietary radiators utilizing black body radiation for heat dispersion, requiring only 1/4 the space needed for heat collection.
  • Biggest Risk: The CEO, Philip Johnson, identified launch costs as the primary risk, making the company a bet on the success and frequent launches of SpaceX's Starship.
  • Markets: Two markets were identified: processing data in space and sending results back to Earth, and in-orbit compute for inter-satellite communication via lasers (e.g., for Starlink array movements).

YC Arena: Learning from Startup Pitches

A new suite of online games called YC Arena, created by a student (MHE 100), allows users to evaluate historical Y Combinator pitches.

• YC Partner Simulator: One game involves watching actual YC application videos and deciding whether the startup was accepted or rejected.
• YC Acceptance Criteria: Jason shared insights into YC's preferences:
  • Team: Three co-founders, ideally two developers, "hackers" over "idea people" or "pitchy people."
  • Context: Historical factors like YC's investment focus (e.g., India, SaaS in 2018) influence decisions.
• Critique of YC: Jason noted a growing "anti-YC vibe" among younger generations, viewing YC as "establishment." He criticized YC's advice to "apply again in six months" as unhelpful for founders and highlighted that YC's 1% acceptance rate means they primarily look for young, family-free developers, which naturally skews male. He also touched on how DEI stats were "gamed" by focusing on the presence of a female founder, often in non-developer roles.
• Launch Accelerator's Approach: In contrast, Jason's Launch Accelerator aims for 50,000 applications/year and 200 investments/year across three programs: Founder University (very early stage), Accelerator (product-ready, some customers), and Syndicate (seed to Series A direct investments). The focus is on "product velocity, deep understanding of customers, a kick-ass team," and a viable business model.

Figure Robotics: Progress and Funding Controversies

Figure, the AI robotics company, was discussed following its $1 billion funding round at a $39 billion valuation.

• BMW Partnership: A new video showed Figure's robot performing repetitive work on a BMW body shop production line, picking up and loading metal parts for welding. This confirmed the robot's practical application, addressing previous skepticism about its "live in production" status. The robots are working 10-hour shifts.
• Funding Model Criticism: The large funding rounds without a fully commercialized product in market led to questions about the funding model. The CEO, Brett Adcock, has used a "different pool of investors" and Special Purpose Vehicles (SPVs).
  • SPV Mechanics: These SPVs often charge a 10% fee on the invested money (e.g., $1 million fee on a $10 million investment) rather than taking "carry" (a percentage of future profits).
  • VC Tension: This model has created tension with traditional VCs who prefer to invest at reasonable prices for 10-20x returns and view SPVs as "transactional" and run by "hustlers" who cater to "dentist crowd" retail investors seeking exposure to "hot" names like Figure, SpaceX, or OpenAI, even at high valuations.
  • Entry Price Matters: It was argued that at such high valuations, the returns might not significantly outperform low-cost index funds (like the MAG 7) over the long term, especially given the additional fees.
• Runway: Despite the high burn rate (potentially $100 million/year), the $1 billion raise provides Figure with a substantial runway (e.g., 10 years), allowing them to develop their technology without immediate pressure.

Onshoring Manufacturing: The Sharpie Case Study

An article from the Wall Street Journal highlighted how Sharpie, a Newell Brands product, found a way to manufacture its pens more cheaply in the US at its Maysville, Tennessee facility.

• Project Origin: This initiative began in 2018.
• Key Factors: The cost reduction was attributed to:
  • Investment in Robotics: Automating manufacturing processes.
  • Upskilling Staff: Training the workforce to operate and maintain the new robotic systems, leading to lower turnover and higher efficiency.
• Broader Implications: This case demonstrates that with "foresight, time, and investment," manufacturing can be cost-effective in the US, challenging the perception that it's always cheaper overseas.
• Profit Margins: The discussion contrasted this with the low profit margins of contract manufacturers like Foxconn, which reported gross profit margins of 6.3% and net profit margins of 2.5%, akin to grocery store margins, despite the complexity of their work. This highlights that while manufacturing happens globally, the majority of profits (e.g., for Apple) remain with the design and brand companies in the US.
• Changing Workforce: The changing demographics of the Chinese workforce (past peak working-age population) were noted as a factor driving increased investment in robotics there, not just for efficiency but also to compensate for fewer available workers.

Deloitte's AI "Slop" and the Future of Consulting

Deloitte Australia faced criticism for a report, for which they charged A$440,000, that was found to contain "classic generative AI mistakes," including citing non-existent sources.

• Criticism: Labor Senator Deborah O'Neal famously stated that Deloitte had a "human intelligence problem" and suggested clients might be "better off signing up for a ChatGPT subscription."
• Impact on Consulting: This incident underscores the fear among consultants of being replaced by AI chatbots.
• Evolving Role: The hosts argued that consulting firms won't disappear but will need to adapt. AI tools will provide initial information, but consultants' value will shift to "first-person interviews, actual research," and critical verification.
• "Show Your Work": The incident emphasized the need for individuals, from students to professionals, to "show their work" and not simply publish AI-generated content without critical review and improvement. The analogy was made to anti-plagiarism software in colleges that prevents cutting and pasting, forcing original thought.
• AI as an Augmentation Tool: AI should be used to enable more and better work, not to replace effort.

Next Visit AI: A Startup Pitch Competition

Ryan Yanelli, CTO and co-founder of Next Visit AI, a participant in Launch Accelerator's 35th cohort (after attending Founder University), pitched his company.

• Problem: Doctor burnout and clinical errors caused by excessive time spent on charting (over three hours/day).
• Solution: Next Visit AI listens to patient sessions, builds clinical data in real-time with deep insights, and finishes charts for review and billing when the patient leaves. It is fast, EHR ready, and HIPAA compliant.
• Results: Dr. Rathor, a user, increased patient load from 16 to 24/day, saved time, and saw a 30% revenue increase.
• Traction: Since launch, Next Visit AI has gained 311 users, 68 paying customers, $9,000 MRR, a 1.6% churn rate, 24% conversion rate, and a "near-perfect" NPS score.
• Metrics: CAC of $189, LTV of $1,700, and ARPU of $133/month.
• Market: Starting with behavioral health in the US, a $2 billion TAM, aiming for 5% market share (60,000 customers) to reach $100 million ARR.
• Competitive Advantage: Positioned as a complete platform providing real-time clinical decision support and accurate data, not just a scribe service.
• Team: Ryan (full-stack engineer with 15 years experience) and Dr. Rafi (psychiatrist with 15 years patient care experience), demonstrating strong founder product fit.
• Feedback:
  • Undercharging: The hosts strongly suggested Next Visit AI is significantly undercharging, given the value provided (e.g., enabling an extra $1,500/day in billings for doctors who make $500k/year). They recommended higher pricing, potentially consumption-based, and direct founder onboarding to secure credit card details immediately after demos.
  • Focus Areas: Obsessively focus on product and customer, studying churn and engagement data to understand user behavior and optimize the product experience (e.g., ideal device form factor).
  • Presentation: Ryan was praised for his "perfect pitch" and natural, passionate presentation style.

---

Conclusion

The episode covered a wide array of topics reflecting the dynamic and often unconventional nature of the current tech and startup landscape. From OpenAI's massive, non-traditional deals to secure AI compute, highlighting the ongoing "compute drought" and the potential for market oversupply (Jevons Paradox), to Tesla's innovative engineering for its next-gen Roadster, the discussion underscored the relentless pursuit of technological advancement. The vision of space-based data centers and the practical challenges of onshoring manufacturing with robotics showcased how fundamental industries are being reimagined. Critiques of traditional VC funding models and the ethical implications of AI in professional services (Deloitte's "AI slop") provided a reality check on the hype. Finally, the pitch from Next Visit AI exemplified the entrepreneurial spirit, demonstrating how focused innovation can address real-world problems like doctor burnout, while also offering actionable insights into startup growth and pricing strategies. The overarching theme is one of rapid change, where traditional rules are being rewritten, and both immense opportunities and significant risks are emerging across various sectors.