Overview of Big Technology Podcast Friday Edition
Alex Kantrowitz and Ranjan Roy spend this episode unpacking three major tech stories: the selloff in AI stocks after the midterm election fallout, Anthropic’s rapid model improvements and the implications for its planned IPO, and whether prediction markets are actually better than pollsters at forecasting elections. The throughline is that AI’s future is still extremely promising, but its business model, infrastructure needs, and public perception are all becoming much more constrained and volatile.
AI Stocks, Data Centers, and the Election Fallout
What happened
- The market reacted sharply to election results and the broader backlash against AI/data centers.
- More than 30 jurisdictions reportedly passed bans or restrictions on data centers.
- A governor in Texas who was seen as relatively pro–data center lost to a Democrat more hostile to the buildout.
Why investors got nervous
- The episode argues that the AI trade depends heavily on a narrow set of assumptions:
- massive data center expansion
- abundant power and land
- smooth execution by a few dominant companies
- If local and state governments slow that buildout, the market’s valuation story weakens.
The key debate
- One side: This is a real constraint and could materially slow AI infrastructure expansion.
- The other side: It may be more of a negotiation than a shutdown. Data centers may still get built, just more slowly and with more local pushback.
- The hosts agree that the buildout probably won’t disappear, but it likely won’t happen at the speed or scale markets had priced in.
What This Means for the AI Trade
Main takeaway
- AI is still a major technology with huge long-term potential.
- But the market is pricing in a very specific, very aggressive path:
- rapid capex growth
- large-scale data center deployment
- continued explosive revenue growth for a small number of firms
Why the selloff matters
- The AI trade is unusually narrative-driven.
- When one piece of the story breaks — especially the data center / compute story — the whole valuation framework becomes unstable.
- The hosts suggest this may slow the pace of AI expansion rather than end it.
Important nuance
- They do not think AI itself is suddenly bearish.
- They do think the economics of the trade may need to adjust to a slower, messier rollout.
Anthropic, Compute Constraints, and the IPO Question
The compute bottleneck
- The episode highlights a practical tension:
- current product inference uses compute
- training the next frontier model uses compute
- safety/security work also uses compute
- If data centers can’t be built as quickly as expected, companies may have to choose which of those buckets gets deprioritized.
Potential consequences
- Slower model training could affect Anthropic’s long-term frontier-model narrative.
- Slower inference could hurt current revenue and user experience.
- Safety/research tradeoffs may also emerge.
Could this actually help Anthropic?
- The hosts float an ironic possibility: if a compute commitment can’t be fulfilled because a project is blocked or delayed, Anthropic may have an easier time stepping back from some of its spending obligations.
- They describe this as a “popsicles in a broken freezer” kind of legal argument — funny, but also illustrative of how messy the commitments may be.
Anthropic’s IPO: Timing, Valuation, and Risk
What’s happening
- The Wall Street Journal reported skepticism about whether Anthropic can justify a $1.5 trillion valuation.
- The hosts note that valuation expectations appear to be drifting downward from earlier levels.
Their view
- They think Anthropic still probably goes public, and soon.
- But they believe the valuation will likely be lower than the most bullish early projections.
Why go out now?
- Waiting would likely mean:
- slower growth comparisons
- stronger competition from OpenAI
- more uncertainty around data center expansion
- a potentially weaker market appetite for rich AI valuations
- Their view is that Anthropic’s momentum is strong enough to support an IPO, but the company may not want to wait for the narrative to deteriorate further.
Bottom line
- The episode’s rough forecast: Anthropic likely still goes public, but not at the previously imagined peak valuation.
- The hosts seem to land around a valuation that is still very high, but meaningfully below the most optimistic scenarios.
Anthropic’s R&D Speed: Are Models Accelerating Too Fast?
Meter’s report
- A nonprofit that studies autonomous AI systems reported that Anthropic’s model-improvement pace may have roughly doubled.
- The report suggests most of the acceleration comes from Anthropic using AI internally to speed up research.
Reactions
- Ranjan is skeptical of how much useful transparency these reports provide.
- His concern is less about whether progress is real — he believes it is — and more about whether the reporting actually helps society respond to the risks.
Key point
- The episode agrees that frontier AI is improving quickly.
- But they question whether the public reporting around that progress is doing anything meaningful beyond feeding fear or helping companies shape their narrative.
Prediction Markets vs. Pollsters
The debate
- Election night brought a bigger mainstream spotlight to prediction markets like Kalshi.
- Markets often moved faster than TV networks in calling races.
Two competing interpretations
- Prediction-market bull case:
Markets aggregate distributed information quickly and may be better at real-time forecasting than traditional media. - Skeptical view:
They may simply be following pollsters and statistical models, not discovering new information.
The hosts’ position
- They think the markets are useful, but they are not replacing polling.
- Prediction markets may be:
- good at translating public information into prices
- fast at reflecting consensus
- less about original insight than critics claim
Concerns
- Both hosts remain uneasy about the gamification aspect.
- They do not love the idea of politics becoming another betting product, even if the data is informative.
Key Takeaways
- AI stock volatility is being driven by infrastructure limits, not just earnings.
- Data center opposition may slow AI growth, but probably won’t stop it.
- Anthropic’s IPO still looks likely, but the valuation story is weaker than before.
- Frontier AI progress is still moving quickly, even if the public reporting around it feels opaque.
- Prediction markets are useful signals, but they may not be as independent from pollsters as their supporters claim.
Final Thought
This episode frames the current AI moment as a transition from pure hype to hard constraints. The technology is still advancing rapidly, but politics, power, land, public backlash, and market skepticism are all starting to shape how fast that future can actually arrive.