Overview of Big Technology Podcast with Alex Kantrowitz
In this Friday news roundup, Alex Kantrowitz and Ranjan Roy unpack a rough week for Big Tech markets, the surprising financial signal buried in Anthropic’s S-1 filing, and the accelerating price war in frontier AI models. The central themes are clear: Wall Street is getting nervous about AI-driven capital spending, Google’s search business is evolving in ways that may weaken its old traffic economics, Anthropic appears far healthier than many expected, and LLM prices are falling fast as model quality and infrastructure improve.
Big Tech’s AI Spending Crackdown
Markets are punishing hyperscalers
- Google, Microsoft, Amazon, and Meta all reported earnings amid intense scrutiny of AI capex.
- The big takeaway: investors are increasingly worried less about “AI spending” itself and more about whether the core businesses behind that spending can still support the outlays.
- The discussion framed this as a potential transition point where these companies are betting on becoming something different:
- cloud/compute providers,
- AI infrastructure platforms,
- or in Google’s case, something less dependent on classic search ads.
Google is the clearest pressure point
- Google still posted strong results:
- Revenue: up around 25%
- Operating profit: up around 30%
- Operating margin: still around 30%
- But search growth came in below expectations:
- search revenue was about $63.3B
- growth was 12%, below the expected 15%
- The hosts argued the issue is not just spending, but whether Google’s core search model can remain as valuable when users increasingly interact through AI mode.
Search behavior is changing
- Google said AI Mode rose from about 2% to 8.9% of searches.
- The conversation focused on a bigger strategic question:
- If users increasingly get answers directly in AI mode,
- do they stop clicking out to websites?
- That would weaken the old bargain of the web:
- websites let Google crawl them in exchange for traffic,
- but AI answers may break that traffic loop.
Reddit’s warning signs
- Reddit CEO Steve Huffman’s comments were highlighted as evidence of the shift:
- Google is still crawling content aggressively,
- but sending less referral traffic than before.
- The broader implication: Google may be extracting information from the open web without sending as much value back.
Amazon, Meta, and the Changing Ad Landscape
Amazon’s cloud and retail mix
- Amazon Web Services showed strong growth:
- AWS revenue: about $46B
- growth: roughly 39%
- But the retail side is also evolving:
- groceries and everyday essentials made up about half of units sold on the retail site.
- The hosts noted that Amazon is increasingly both:
- a cloud company,
- and a deeply embedded consumer logistics/retail platform.
Closed ecosystems may be winning
- A key argument: Google’s ad model depends on the open web, which AI is disrupting.
- Meta and Amazon, by contrast, operate more closed ad ecosystems:
- ads live inside their own platforms,
- and are less dependent on outbound traffic.
- That may put them in a structurally better position long term, even if both are also making heavy AI bets.
Market valuation signals
- The discussion pointed out that public markets are already expressing skepticism:
- Alphabet down on the year,
- Meta down,
- Amazon lagging the S&P,
- Microsoft roughly flat to slightly positive.
- The takeaway: investors may be saying the biggest AI bets are also the riskiest.
Anthropic’s S-1: Stronger, Richer, and More Complicated Than Expected
The filing looked surprisingly healthy
The S-1 was framed as a Rorschach test:
- If you’re bullish, it shows a fast-growing business.
- If you’re bearish, it shows huge future obligations and uncertain economics.
Key positives:
- Customers spending $100K+ grew from about 1,500 in Dec. 2024 to 11,900 in June.
- Anthropic reportedly has $110B+ cash on hand.
- The company’s revenue efficiency looked strong:
- it now makes $2.08 for every $1 of compute spent
- versus $0.40 before
- It is generating about $6.6M per employee.
But the real unit economics are still unclear
The hosts stressed that the S-1 still doesn’t answer the biggest question:
- What is Anthropic actually paying for compute at market rates?
- How much of its efficiency reflects discounted or subsidized infrastructure from Amazon and Google?
Discounted compute may be distorting the picture
- Amazon and Google have invested heavily in Anthropic and may also be offering discounted compute.
- That means the S-1’s margins may not reflect the true future economics once those sweetheart terms change.
- The hosts argued investors should push for more disclosure around:
- what Anthropic actually paid,
- how much was discounted,
- and how those costs compare to market rates.
The wealth creation is enormous
A major segment focused on how many people inside Anthropic could become wealthy if the company lists at a very high valuation:
- At a hypothetical $2T valuation, employee equity could be worth hundreds of billions.
- One estimate discussed:
- roughly $331B in employee options/RSUs
- average value in the tens of millions per employee
- Employees who joined earlier could potentially be worth:
- $100M+ for many early hires
- even newer hires could see massive paper gains
The political optics are wild
- The hosts noted the irony that Anthropic’s CEO has warned that AI could automate a huge share of white-collar jobs.
- At the same time, the company may mint a large number of ultra-wealthy employees.
- That tension—massive job disruption alongside enormous private wealth creation—was presented as a ready-made political talking point.
AI Costs Are Falling Fast
LLM price cuts are happening quickly
The episode highlighted a sharp decline in model pricing from both OpenAI and Anthropic:
- price cuts of 40% to 80% in some cases
- evidence that frontier models are becoming more commoditized
- easier model switching for developers using routing tools or abstraction layers
Why prices are falling
The hosts discussed three likely reasons:
- Commodity pressure
- model quality is converging,
- differentiation is harder,
- customers can switch more easily.
- Better infrastructure
- models may genuinely be cheaper to serve now,
- which would support lower prices.
- IPO signaling
- companies may want to show growth and usage momentum before going public.
Bigger implications
- If models get much cheaper, it could weaken the case for endless data center expansion.
- That matters because the AI infrastructure boom has become one of the main market narratives supporting Big Tech capex.
- In other words:
- cheaper inference could be good for customers,
- but bad for the economics of the build-out.
Other Notable Moments
The Amazon Bee frontier wearable
- The hosts had a long aside about Amazon’s AI wearable, the Bee.
- Kantrowitz was strongly against a world where conversations are continuously recorded and summarized.
- The concern was social as much as technical:
- if everything can be replayed,
- people will behave like they’re always on stage.
A fun pop-culture detour
- They briefly discussed The Dropout / Theranos material and Nathan Fielder’s project The Rehearsal.
- The main point: the new Fielder work is weird, but in a compelling way, and still worth watching.
Key Takeaways
- Big Tech is getting punished for AI capex when the core business starts to look less secure.
- Google’s search model is changing in ways that may hurt the open-web traffic economy.
- Anthropic looks operationally strong, but its true economics are still cloudy because of possible discounted compute.
- The AI price war is real, and it may be a sign of both commoditization and better infrastructure.
- Public markets are skeptical, but private AI valuations and wealth creation remain enormous.
Final Thought
The episode’s underlying message is that AI is simultaneously making some companies look fragile and others look extraordinarily powerful. Public investors are starting to demand proof that the old business models still work, while private markets are still rewarding the companies betting hardest on the future.