CNBC International Live Sep 18, 2026
With Rajiv Batra, Head of Asia and Co-Head of Global Emerging Markets Equity Strategy at J.P. Morgan
The companies renting out AI compute have raised capital spending by 125%, while the hardware available to meet that spending has risen 70%. On Rajiv Batra's numbers, that gap is why he still reads the economics of the AI build-out as profitable.
Equities have fallen since July on rising yields rather than on disappointing profits. Batra's view, which he said may sound contrarian, is that higher yields and higher equity prices can run together, because the index is no longer built out of the companies that rates hurt.
"Market is getting scared about the yields and trying to draw the pictures from the historic books over there."
Batra heads Asia equity strategy at J.P. Morgan and co-heads its global emerging markets equity strategy, and his argument rests on what the benchmark is made of rather than on a forecast for rates.
The full segment is covered here so you can skip it.
Here are the 5 numbers that matter.
Key Takeaways
The fall since July is a derating, not an earnings problem — Batra called profits "fantabulous" across the US, Japan and emerging markets
Today's benchmarks are less rate-sensitive, weighted toward AI services and healthcare, which is why he says high yields and high equities can coexist
Hyperscaler capital spending is up 125% against a 70% rise in hardware, leaving a demand-supply gap
Hyperscaler returns on that spending are still running around 25% to 42%
Token prices fell while token use rose 10x to 20x, and the revenue base has reached $260–280B
1. Derating, Not Bad Earnings
Asked whether a higher-yield environment derails the case for Japan and for the region, Batra agreed with the premise and then reframed what has actually happened to prices since July. Global equities have fallen because the multiple came down, not because profits disappointed. He called earnings "fantabulous" across the United States, Japan and the emerging market world.
The market is reacting to yields, not results
Market is getting scared about the yields and trying to draw the pictures from the historic books over there.
Rajiv Batra
2. High Yields and High Stocks
His contrarian claim follows from that. What decides whether rising yields hurt an index is what the index contains.
Yields and equities can rise together
But what we believe and may sound contrarian that higher yields and higher equities can coexist in this world over year and the important part is we need to look at the compositions of the benchmark.
Rajiv Batra
The benchmark is now weighted away from rate risk
Today my compositions of the benchmark are less rate sensitive and more focuses on AI services and healthcare who are insensitive to what is happening in the rate market overall.
Rajiv Batra
He does not expect yields to fall back. Deficits and debt-to-GDP are structural, he said, and yields have moved in one direction since the pandemic with no mean reversion. His point is that this does not require equities to roll over.
So he expects yields to stay high for a long time
So taking that into account because of the structural concern be the deficit debt to GDP yields can remain higher for sustainable period
Rajiv Batra
3. Compute's Two Overhangs
The host set out the case that the AI trade has reached another inflection point, and asked how to approach the Asian markets whose companies are competing with the American leaders.
The host listed three reasons to doubt the trade
The other dynamic globally is whether AI and the entire AI tech momentum trade is yet again at another inflection point whether it's because of China competition public opposition to data center buildouts or this entire existential risk to humanity.
A host
And asked where to put money in Asia
How then should we approach Taiwan, South Korea, Japan and China and these AI champions that are really giving their US rivals a run for the money.
A host
Batra granted two overhangs on the business of selling compute: the safety and regulatory questions, and the financing of the build-out.
Safety rules and funding both remain open
So yes, safety and the regulation concerns have started coming on the compute business and that started raising question mark whether the compute business will remain sustainable over there and remember the financing issue is still not off the table yet.
Rajiv Batra
But the economics still work on his numbers
But when we look at the larger scheme of thing as such we still believe the AI economics remain profitable over there.
Rajiv Batra
4. A 125% vs 70% Supply Gap
Batra split the AI economy into three groups and checked each one: the companies that host the compute, the companies that make the hardware, and the companies that build applications and models on top.
The hosts are still earning 25% to 42%
Most importantly when we look at the hosting guys so the hyperscalers earnings or ROI they're still in the zone of around 25 to 42 odd% per se
Rajiv Batra
Their remaining performance obligations, the contracted revenue they have booked but not yet delivered, run into the trillions. Their capital spending is up 125%, while the hardware rising to meet it is up 70%, which leaves demand well ahead of supply. The group he says the market has started to doubt is the third one, the applications and the model developers.
5. Prices Fell, Volumes Rose
The worry about the model companies is that cheaper open-weight and open-source tokens have pushed prices down. Batra's objection is that the market is reading one half of the equation. Token use has risen 10 to 20 times, and the revenue base has reached $260 billion to $280 billion.
The market is watching price and ignoring volume
So we are looking at pricing we are not looking at volumes yet.
Rajiv Batra
What he cannot answer yet is what those revenues earn, because the companies are private.
Margins stay unknown until the listings come
The questions are there on margins which hopefully once the public issuances and all that come we will come to know what the net margins or gross margins are happening around
Rajiv Batra
Bonus Insights
The employment case against AI has not shown up in the data
Batra said the argument about whether AI is good for humanity will keep coming back in a cycle like this one, but that so far there has been "no job displacement" and no wave of firings. On the payroll prints he reads, the global picture is "net hiring at a global level".
Batra's bottom line is that the derating since July is about yields rather than earnings, and that an index weighted toward AI services and healthcare can keep rising while yields stay high.
Products, Companies & Tools Mentioned
J.P. Morgan (Batra's firm, and the source of the benchmark-composition argument, the hyperscaler return range and the token-volume figures he cited)
Watch the full episode:
More Rajiv Batra interviews and podcast appearances
If this was worth your time, send it to someone who has to have a view on this.
Get the latest market chatter as it happens: