Overview of Big Technology Podcast
In this Friday edition, Alex Kantrowitz and Ranjan Roy dig into three big AI stories: how Big Tech may be hiding trillions in AI infrastructure obligations off its balance sheets, what newly leaked revenue numbers suggest about Anthropic’s rapid rise versus OpenAI’s growing pains, and why travel may be one of the best real-world evaluations of AI products. The conversation is skeptical but not cynical: they argue the AI buildout is real, but the financing, reporting, and competitive dynamics around it are getting increasingly extreme.
Big Tech’s Hidden AI Spending
What the Wall Street Journal story uncovered
- The hosts focus on a WSJ report arguing that Big Tech’s AI spending is about $3 trillion higher than it appears on the surface.
- The key issue: much of the spending is being structured as off-balance-sheet commitments, not traditional capital expenditures.
- The companies cited include Alphabet, Meta, Oracle, and others, with obligations tied to data centers, leases, and financing vehicles.
How the structuring works
- Rather than directly funding data centers with cash, companies use external financing structures:
- private capital vehicles,
- joint ventures,
- bond financing,
- collateralized chip/data-center arrangements.
- Example discussed:
- Meta’s Hyperion data center in Louisiana.
- The campus and its financing sit largely outside Meta’s balance sheet.
- Meta is the tenant/partner and has a long-term lease/guarantee arrangement, but the debt itself sits elsewhere.
Why they do it
- The obvious incentive: preserve cash and keep reported financials cleaner.
- The less obvious but equally important incentive: avoid shocking Wall Street with the full scale of AI-related commitments.
- The hosts argue this is logical from a company perspective, but potentially misleading for investors trying to assess real leverage and risk.
Why the hosts are concerned
- These companies are usually known for fortress balance sheets and massive cash generation.
- Some are now showing signs of negative free cash flow or much heavier obligations than investors may realize.
- The scale is so large that it raises questions about:
- systemic risk,
- leverage concentration,
- whether AI demand will justify the buildout.
Anthropic vs. OpenAI: Revenue, Momentum, and IPO Readiness
Anthropic’s surge
- Bloomberg-reported numbers suggest Anthropic’s annualized revenue has reached roughly $65 billion, with a quarterly figure of $11.5 billion.
- The hosts see Anthropic as the clearer winner in the current moment:
- faster growth,
- stronger momentum,
- better positioning for an IPO.
OpenAI’s challenges
- OpenAI’s reported revenue growth is still strong, but its pace appears slower relative to Anthropic.
- The transcript references:
- 18% quarter-over-quarter revenue growth,
- deeper losses,
- leadership turnover,
- and a sense that the company is behind Anthropic in enterprise/coding momentum.
Leadership churn at OpenAI
- Notable departures mentioned:
- Denise Dresser (chief revenue officer),
- Brad Lightcap,
- Fidji Simo.
- The hosts interpret this as evidence of organizational turbulence, especially as OpenAI tries to pivot and scale into enterprise markets.
What they think it means
- Anthropic appears to be pulling ahead for now.
- OpenAI is still huge, but the market is starting to scrutinize:
- execution,
- monetization,
- enterprise strategy,
- and whether its leadership structure can stabilize before an IPO.
The Debate Over AI for Travel Use Cases
Alex’s case for travel as an AI eval
Kantrowitz argues that travel is an ideal use case for AI because it sits in a “Goldilocks zone”:
- important enough to matter,
- but not so critical that one mistake is catastrophic.
His three main points:
- Travel requires synthesizing lots of information
- where to stay,
- what to do,
- how to get around,
- what’s open, what’s not.
- Travel is a dynamic, multi-step planning task
- which makes it a good proxy for real-world work.
- It’s low-cost enough to experiment with
- so if AI performs well here, it can build confidence for harder tasks.
Ranjan’s initial skepticism — and eventual agreement
- Roy initially pushes back, but after recent travel experience, he largely agrees.
- He says AI now handles complex travel logistics far better than before:
- itinerary planning,
- hotel and transport coordination,
- reminders,
- pulling context from email/calendar,
- generating structured planning docs.
A real example
- Alex describes using AI to plan a trip that led him to Denia, Spain, based on detailed requirements like:
- Airbnb with a pool,
- sandy beach,
- shallow snorkeling,
- value constraints.
- Ranjan describes using AI for a multi-island trip in Indonesia and Dubai, with daily automated briefs and logistical support.
Why this matters
- They agree that travel may be one of the best “normie” evaluations of AI:
- easy to test,
- realistic,
- useful,
- and revealing about product quality.
- It’s a strong indicator of whether AI can reliably manage messy, changing, human workflows.
Main Takeaways
- Big Tech’s AI buildout is much larger and more leveraged than public CapEx figures suggest.
- Off-balance-sheet financing is helping companies hide or defer the true scale of AI infrastructure commitments.
- Anthropic currently looks like the stronger AI company financially and operationally, while OpenAI is showing signs of internal strain.
- Travel is a surprisingly strong benchmark for AI utility, because it combines complexity, planning, and real stakes without being mission-critical.
- The broader theme: AI may be transformative, but the financing and execution risks are becoming harder to ignore.
Notable Insights
- “If the data center story works, then it is not a cash outlet.”
- “It’s a call option on AGI.”
- “Travel is in the Goldilocks zone of important, but not so important tasks.”
- “Anthropic is putting some distance between itself and OpenAI.”
Bottom Line
The episode paints a picture of an AI industry that is still booming, but increasingly complicated underneath the headline numbers. Big Tech’s hidden commitments suggest the buildout is far more aggressive than reported; Anthropic appears to be accelerating faster than OpenAI; and travel emerges as a practical, high-signal way to judge whether AI is actually getting better at real-world work.