[Free Post] Jensen’s law, CoreWeave is the best, AI Search is coming in hot, and a smooth transition to streaming
A menagerie of AI-related things, with a segue into streaming, YouTube, Meta, and India
ChatGPT says:
Doubling Down on Moore’s Law: NVIDIA redefines growth, crushing the classic "every two-year" rule with an astonishing 2.3x annual increase in compute power.
CoreWeave Shakes Up the Cloud: Despite a shaky IPO, CoreWeave is touted as the GPU cloud leader, outpacing giants like AWS and Google.
GenAI Search Shake-up: Legacy search is crumbling under GenAI's prowess, with savvy banks leveraging AI-driven referrals for major competitive gains.
Undervalued Giants: YouTube and Instagram, despite their massive user engagement, remain surprisingly under-monetized in the digital ad space.
India’s Digital Leap: As mobile dominates, digital media earnings in India surpass TV for the first time, hinting at a vast monetization frontier ahead.
👉👉👉Reminder to sign up for the Weekly Recap only, if daily emails is too much. Find me on twitter, for more fun. Jensen’s law, CoreWeave is the best, AI Search is coming in hot, and a smooth transition to streaming
Friday fun day, so this will be brief (and random)
Jensen’s Law
According to EpochAI, compute power from NVIDIA chips doubles every 10 months:
Total compute attributable to Nvidia’s chips has grown ~2.3x per year since 2019.
A little while back, Random Walk flagged some (self-serving) speculation from TSMC that computing needs were growing more quickly than chip-density could keep up with. Perhaps that’s still true, but for now, Nvidia has doubled the pace of Moore’s Law, by doubling+ every year, instead of every two years.
As I wrote then, “Moore’s laws were meant to be broken.” Well, Moore’s Law has become Jensen’s Law.
CoreWeave is the best!
CoreWeave’s IPO has gotten off to a rocky start, but according to some folks, it really is the best GPU Cloud offering:
CoreWeave sits at the top for being the best at the most things.
Good for CoreWeave, and certainly I’ll have to take SemiAnalysis’ word for it, but the other thing that stands out are the relative rankings of the big dawgs: AWS and Google Cloud in 3rd and 4th tiers respectively.
Pressure on Sundar continues to mount.
Search is toast (and so are the firms that don’t learn GenAI search optimization)
Just another entry in the mysterious decline of Google’s unbeatable search monopoly, which the highest authority of the land has declared unbeatable. Curious.
GenAI search is driving an increasing share of very valuable traffic, and the firms that seem to have figured out how to gain priority in its ranking, are making outsized gains.
In this case, take a look at webtraffic for UK financial institutions, where high intention customer leads are worth their weight in gold (because LTV for financial services customers is so high):
Referral traffic for challenger banks has grown, while legacy incumbent institutions have seen substantial declines.
Referrals have accounted for ~11% of site traffic for the financial sector in the UK, which is ~2x the next closest sector. In other words, referral traffic is very important to UK banks. If it’s changing, that’s a pretty big deal.
Do we know for sure that it’s GenAI causing the shift? We do not, but there are some clues.
The share of referrals coming from AI search is ~2x higher for challenger banks than legacy banks:
That’s 3 per every 1,000 for Revolut and co, and ~1.5/1,000 for the incumbents.
Challengers are getting an outsized share of AI search leads. Is that intentional or luck? Idk, but someone is going to figure out how to do that intentionally (and they already are).
Google better turn on those monopoly powers fast, amiright?
YouTube and Meta are under-monetized
Just to continue to riff on a theme . . . while Google may have some headwinds on other fronts, YouTube is definitely not one of them.
The shift to streaming video and decentralized content-creation, and all it’s various forms and glory, is still just getting started, and under-appreciated, both in its economic impact, but it’s cultural impact, as well.1 We are what we eat, when it comes to content, and the generational step-changes in use portend generational step-changes in culture.
Basically, people watch a lot of YouTube, but it turns out they watch a lot of Instagram too:
Avg time per day on Insta has grown 25% since 2022 (and 16% for YouTube).
At the same time as Insta and YouTube have risen, TikTok’s (and Facebook’s) usage have actually fallen. YouTube and Insta, two of the greatest acquisitions of all time. My goodness.
When it comes to paid-channels, Instagram is running away from the field:
A staggering 80% of surveyed marketers are using Insta for influencer marketing.
80% (up from 70% in 2023) is incredible, and presumably that’s just money going directly to influencers (and not the platform ads).
YouTube, by contrast, is still under 50%, which means there is plenty of runway to penetrate the ecosystem (and YouTube’s creator payouts are relatively high).
In terms of who would “win” the short-form video wars, if TikTok gets the boot? It’s a race between Zuck and Zuck:
Facebook Reels and Instagram Reels are 1-2 for “next major short-form video platform.”
It’s good to be Zuck.
Big in India
And finally, because Random Walk has gotten a kick out of this before, Zuck has another remarkable advantage in his bag o’ tricks: India.
India is a fast(er)-growing, “digitally native” country, in that it joined the internet co-terminus with mobile, and so mobile and “internet usage” are basically one and the same. As the rest of India comes online (and there is still ~40% left to go), that’s a whole lot more mobile users—and that’s in addition to the ~320 million Indians already on facebook (mostly WhatsApp).
Monetization of those users is currently paltry, but there is (apparently) plenty of money to be made in India with digital media:
Digital media has surpassed TV in India as a source of media revenue.
Indians collectively spent 1.1 trillion hours staring at their smartphones in 2024 . . .
On average, they spent five hours daily on the mobile screen, nearly 70% of it devoted to social media platforms, gaming, and videos . . .
This has made digital channels the single largest segment of India’s 2.5 trillion rupees ($29.1 billion) media and entertainment industry in 2024, overtaking television for the first time.
1.1 trillion hours of phone time, and only ~half the country has phones.
Incredible stuff.
Other links
Brazil’s “Egg King” shells out $1B to acquire major US producer. M&A is back.
Taxpayers pay twice to cover the same people with Medicaid. Taking this at face value, if someone enrolls in one state, and then another, both states get paid for the care. A good reminder that whatever mistakes DOGE makes, the margin for error is extremely wide, given the bar for incompetence and waste that’s already been set. The status quo deserves no deference—quite the opposite.
Real Estate funds took a record 21.8 months to close in 2024. Fundraising is tough out there, especially in real estate. Everyone wants to lend, but equity? No sir. And why fund invest when you can just do deal-by-deal co-invest?
Previously, on Random Walk
Volatility begets volatility, and other sundries
Vol begets vol (and the analysts get it wrong, a lot)
Healthcare makes all the jobs . . . but maybe manufacturing too? Other labor market desiderata
healthcare jobs replaced all the manufacturing jobs (much to the benefit of women and aides)
Random Walk is an idea company dedicated to the discovery of idea alpha. Find differentiated data, perspectives and people, and keep your information mix lively. A foolish consistency is the hobgoblin of small minds. Fight the Great Idea Stagnation. Join Random Walk. Follow me on twitter. Follow me on substack:
If RFK Jr. actually bans pharma from advertising, Cable is toast sooner than expected (another unbeatable monopoly, on the highest authority, I might add).