The Compound and Friends Sep 18, 2026 55m 37m saved
With Dan Ives, partner and senior managing director at Yorkville Ives and Co. and previously eight years as global head of technology research at Wedbush · Tom Lee, co-founder, head of research and Chief Investment Officer at Fundstrat Capital
Orders for AI chips outrun the supply that can be made by 13 to one, Dan Ives told a live audience, and he does not expect the market to reach equilibrium until early 2029.
The two guests were on stage the day before a Federal Reserve meeting that the futures market had priced for a 25 basis point hike, with semiconductors 19% off their highs and the heads of the largest AI labs publicly calling for a slowdown. Both read all of that as bullish.
"Demand to supply right now for chips is called 13 to 1."
Ives spent eight years as global head of technology research at Wedbush before co-founding Yorkville Ives and Co. in July. Tom Lee co-founded Fundstrat, which marked its twelfth anniversary the day of the recording, and runs the Granny Shots exchange-traded funds.
The full episode is covered here so you can skip it. 55 minutes of audio, 18 minutes of reading.
Here are the 15 insights that matter.
Key Takeaways
Lee said earnings are up 25% this year against a market up 10%, so the multiple has fallen
Ives put chip orders at 13 for every 1 chip that can be produced, with no equilibrium before early 2029
Every dollar of capital spending carries a $5 to $6 multiplier across the rest of technology, on his numbers
Lee's list of what kills a bull market is short: a genuine bubble, or the Fed, which he blames 80% of the time
Nvidia trades at 16 times earnings against 50 for Costco, which Lee calls the Apple app-store situation again
800 billion a year of capital spending going to 1.1 trillion is why Lee reads the safety push as positioning
Robots would be a third productive input after labor and capital, which is the only version that grows the economy
Ives puts the odds of Tesla and SpaceX merging by the end of next year above 80%
Consumer discretionary is shrinking because it is the catchall sector, not because spending has changed
Both called this a 1997 moment rather than 1999, with software the downstream beneficiary of AI
1. The Rally Starts Tomorrow
Asked where the market stood going into the Fed decision, with the biggest capital-spending wave in dollar terms in American history running alongside an open argument about AI safety, Tom Lee started with sentiment.
The wall of worry is the setup
There is a wall of worry that you're describing and prices have reacted, especially the rate hike because obviously nobody likes a Fed to tighten.
Tom Lee
His argument is that pessimism is high, earnings are in good shape, and a delivered hike, which he put at 90% probability, takes future hikes off the table because the market is relieved the Fed acted.
He gave it a time
So I think the probability of a massive rally starting tomorrow is really high.
Tom Lee
Pressed on the hour, the room settled on 2:15, and one of the hosts amended it to 2:16 to allow for the statement being read.
The hosts then listed what the market has absorbed this year without breaking: the Strait of Hormuz closed and still closed with oil spiking, tariffs, persistent inflation, an unwind in the AI trade, the 10-year Treasury at 5%, a hiking Fed, a shut housing market, and now the frontier labs discussing a slowdown.
And yet
The S&P is 2.6% off of its all-time high.
A host
Including the part everyone assumed was propping it up
Semis, which were I guess the only thing holding up the market, not true.
A host
Semiconductors are in a drawdown of about 19%, which the hosts put to the guests as evidence that the market is bulletproof rather than fragile.
2. Earnings Up 25%, Price Up 10
Lee's answer to the bulletproof question was arithmetic rather than sentiment.
The multiple has been falling all year
This year earnings are up 25% and the market's only up 10. So the market got cheaper.
Tom Lee
His explanation is that companies adapted to each shock by cutting costs, and that the discipline is the reason to own them.
Which he frames as a management story
We're trusting our money with these CEOs that are proven that they're blue chip because they're not just sitting and telling the board everything's fine.
Tom Lee
They are in crisis mode and cutting costs all the time, he said, which is what produced the earnings.
3. 13 Orders Per Chip
Asked what he tells people who point at contracting valuations on the most profitable companies in American history, Ives started with the competitive position.
The gap with China has reversed
Well, for the first time in 30 years, the US is way ahead of China when it comes to tech.
Dan Ives
And the supply position is extreme
Demand to supply right now for chips is called 13 to 1.
Dan Ives
Which he spelled out
There are 13 orders for every one chip that can be produced.
Dan Ives
With no balance in sight
You're not going to have true equilibrium probably till early 2029 at this pace.
Dan Ives
He said that is what the bears arguing about valuation from a New York office building do not see, and that he and his team see it in Taiwan and Korea. His transmission mechanism runs outward from the chips.
Every capex dollar multiplies
And for every dollar spent on capex, there's a five to $6 multiplier across the rest of tech.
Dan Ives
That, he said, is why software, infrastructure and cybersecurity follow.
Which leads to his estimate gap
Investors are underestimating earnings by probably 25 30% next few years.
Dan Ives
Pushed on whether such a run of upside surprises can continue, he named memory as the group at the center of it, pointing at where Micron trades.
His stage of the cycle
We still believe this bull market's young.
Dan Ives
Because of what is still to be built
You have 12 to 2,500 data centers that are going to be built next five to six years.
Dan Ives
Even assuming 10% or 15% get voted down, he said, the innovation boom is only starting.
4. Where the US Leads China
Asked directly how he justifies the claim that the United States is ahead, given open-weight Chinese models and the pace of DeepSeek and Tencent, Ives went layer by layer.
It starts with one chip
There's one chip in the world fueling the AI revolution.
Dan Ives
On semiconductors he put the US two to three years ahead.
Even the third-tier product leads
Their third rate chip is still probably a year ahead of where Huawei is just to keep in perspective.
Dan Ives
He said the hyperscalers he named, Amazon, Alphabet and Microsoft, are not in the same category as China's large technology companies, and that the frontier models are not comparable to Chinese open-source alternatives. He was equally clear about the other direction.
Two areas where China leads
China's ahead of us when it comes to robotics and energy.
Dan Ives
His conclusion is conditional rather than triumphant: withholding chip sales to China, or slowing model development, closes the gap rather than widening it, because China is not slowing down.
5. What Kills a Bull Market
Asked for the most credible bear case he hears, Lee named two.
Only two things end one
Well, I think two things will kill a bull market.
Tom Lee
The first is a genuine bubble, which he said could arrive if the market concluded one AI model had won and capital spending went to zero. The second is policy.
And one of them is far more common
The Fed 80% of the time is the reason a bull market ends.
Tom Lee
He said the Fed would act if it genuinely believed the economy was overheated, inflation pressures were unhinged and a debt bubble was forming. A host then put the debt-issuance case to the table: after roughly 15 years of companies shrinking their share count since the buyback era began around 2012, there are now large initial public offerings and heavy corporate bond issuance crowding out Treasury sales.
The answer was a definition
A debt bubble takes place when there's quote return risk.
Tom Lee
The condition, on that definition, is money cheap enough that bad projects get funded; data-center funding costs are rising instead.
Which is the market doing the work
The market is already putting a break on everything.
Tom Lee
6. The Data Center Backlash
The risk Ives named instead is political, and he was blunter about it than about anything else in the conversation.
His actual worry
Every data center that gets voted down, China wins.
Dan Ives
He said he has met many of the politicians involved and does not want people carrying flip phones and BlackBerrys deciding the country's technology policy. His argument for the other side of the ledger is employment: the jobs come from the build-out, and he contrasted spending time in Taiwan, where he said fabs are built 18 hours a day, with landing at a New York airport and finding a fist fight at the Dunkin' Donuts while the country sits 17th in mathematics.
The hosts laid out the three fronts of the political argument: resource use, water and electricity; the job-loss statements from founders, including the claim that half of white-collar jobs disappear by 2030, which get walked back but not unheard; and the fear of a catastrophic hack.
7. The Anthropic Warning
The specific event behind the episode was a blog post published on the Saturday by Dario Amodei, which the hosts summarized as an emergency warning: things seen in his own lab and others that worry him, a call for outside observers inside the frontier labs, and a call to slow the pace of development. Half the audience already agreed; the other half called it a hoax.
Lee's reaction was that the timing was late rather than suspicious.
Regulation always lags growth
It's overdue actually because every industry that grows always grows ahead of regulation.
Tom Lee
He noted that AI has no self-regulatory organization at all, and that the call comes while communities are organizing against it.
The opposition is already counted
There's like 38 anti-AI groups and there's like moratoriums on data centers.
Tom Lee
So the move makes sense to him
It's smart to say let's self-regulate one and now we're acknowledging that we got issues so we can talk to the communities.
Tom Lee
His precedent is the mobile industry: the panic about radiation from handsets and masts, the attempts to ban cell towers, and the federal legislation that eventually settled it. He noted that phones used to transmit at a full watt and now transmit at 25 milliwatts, and asked what would have happened had the industry been shut down instead.
Pressed on whether he thinks the response is genuine, Lee reframed it as commercial.
The number that explains the positioning
Think of it as it's 800 billion a year is the capex and it's going to go to 1.1.
Tom Lee
He does not think it falls to 400 billion. What the labs want, in his reading, is public acceptance of spending at that scale, and he called it a smart move.
Because the audience is not shareholders
Every CEO knows today they can make public statements to influence politicians and voters, not just shareholders.
Tom Lee
Asked whether it is marketing, Ives allowed that a step on safety was needed and then said the industry made its own problem.
The public relations damage was self-inflicted
Dario himself and a lot of these tech companies created this PR nightmare themselves.
Dan Ives
Tell people they will lose their jobs and their electricity bills will rise, he said, and they do not celebrate a data center appearing nearby. He also named the competitive motive plainly.
His description of regulatory capture
When you get to the penthouse then you stop the elevator so others can't come up.
Dan Ives
He pointed at software stocks as the market's read on it: if the leaders slow, the gap narrows. Asked what China winning actually means, he listed it.
What losing the race would cost
It means on chips, on models, on infrastructure, on robotic all the technology that's going to be built in AI.
Dan Ives
Lee added a second-order worry: whoever controls the models controls the recursive loops people now rely on, which he described as a form of mind control. One of the hosts raised the timing of Mark Zuckerberg releasing a free open-weight model days before the warning; the answer from the table was that competition is broader than the two leading labs, and that startups and large technology companies will close the gap regardless.
Asked whether legislation is likely before year-end, the answer was no. Drafting a law that says AI cannot be that good is not straightforward, and regulators know they are behind, which is why the proposal is for the labs to evaluate each other rather than have government in the middle.
8. Sovereign AI
Ives raised the risk he thinks matters more than regulation, crediting Alex Karp of Palantir for making the same point repeatedly. With hundreds of models coming, the models themselves become commoditized; what does not is the data.
His definition of sovereign AI is a company keeping control of its own data and refusing the model providers access, because handing it over creates a risk to the business itself. Corporations, on that account, will train their own models on data nobody else can reach, which is a problem for general-purpose models. One of the hosts added the logical extension: competition then makes public data less trustworthy.
On Elon Musk endorsing a slowdown, the reading offered was positional. Musk has been consistent on safety, but his own model is behind, and the two leaders are far enough ahead that nothing short of a regulatory intervention closes the gap.
9. Nvidia at 16x
Semiconductors are 19% off their highs while Nvidia keeps reporting extraordinary quarters. Asked whether the group makes new highs in 2026, Lee reached for a precedent.
The comparison is the app store
You're paying 16 times and you're paying 50 times to buy Costco and 48 time to buy Walmart
Tom Lee
Nobody gave Apple credit for the app store either, he said, until the multiple doubled one day; if Nvidia charged a membership fee, its multiple would go to 50. Asked whether the market being that wrong worries him, he pointed at who has not participated.
The reluctance is the opportunity
Bears have called 10 of the last two downturns.
Tom Lee
He said he speaks to fund managers who are underweight or hold nothing in AI because they thought it was a bubble, that calling a structural story a bubble is difficult, and that institutions still have room to buy both AI and the new listings. The wealth being created by the private labs, he noted, reaches a narrow group of people.
10. The SaaS Apocalypse
Six months earlier, the largest software companies in America were in drawdowns of 20% to 60% with nothing wrong in their results, on what the market called the software apocalypse. Asked for an update, Ives was scathing about the original narrative.
He called it fiction at the time
Narratives create the opportunities
Dan Ives
He allowed that Adobe and Intuit have genuine structural questions, but said the idea that the frontier labs would wipe out ServiceNow, Salesforce and Palantir was close to bad comedy. His illustration was cybersecurity: a frontier model announced a security product around the time of the industry's April conference, the stocks were crushed, and people told him the sector was finished.
What the sector actually did
The three largest cyber security stocks, CrowdStrike, Fortinet, Palo Alto are all up 100% plus on the year.
Dan Ives
Lee's version was about how the selling happened.
Software is downstream, not upstream
Investors now understand software stocks are downstream beneficiaries of AI. They're not victims of a boogeyman.
Tom Lee
What has become easy, he said, is pressing a button and taking an allocation to zero, which is the same reflex that showed up when the Fed signals a hike.
11. Robots as a Third Input
Asked whether a genuine robotics trade is what takes the market higher, Lee built the case from national accounts.
Two levers become three
If you add robots, you actually have a third productive output unit that may not consume people or capital.
Tom Lee
The economy is labor and capital, he said, and adding a productive unit that consumes neither means growth without inflation: a Fed could permit 7% growth, and robots that become consumers and taxpayers would help with government deficits. He imagined robots running their own economy with humans taking the dividends, and offered stone masonry coming back so houses can look carved as the everyday version.
The condition that makes it work
If robots are just replacing jobs then the apocalypse is correct.
Tom Lee
Ives put the bigger number on the adjacent category rather than on humanoid robots.
Autonomy is the prize
Physical AI to me is the golden goose.
Dan Ives
He argued physical AI could be a larger capital-spending category than anything so far, and that true autonomy will be among the biggest technology innovations anyone sees.
12. Tesla and SpaceX
Asked which stock is the highest-probability way to own the autonomous future, the answer was Tesla, then SpaceX, then a prediction.
He put a probability on a merger
It's over an 80% chance by the end of next year that Tesla and SpaceX ultimately merge.
Dan Ives
Ives described one company holding both as the golden vision, not only for the data but for the AI technology being built across the two, and said the result would be the largest company in the world.
13. Discretionary Is Shrinking
One host raised a ratio he had discussed the previous week: equal-weight consumer discretionary against the equal-weight S&P 500, making new lows continuously. Lee's answer was that the reader is looking at a classification artifact.
The sector is a residual category
Historically consumer discretionary is what you call the catchall sector.
Tom Lee
A stock is defined as industrial, technology, healthcare or staples, he said, and whatever fits none of those is discretionary. Discretionary spending has not changed; iPhone spending is counted in technology and a warehouse-club trip in staples, so the leftover shrinks. He used Delta as the awkward case, which one of the hosts extended: Delta owns a refinery and trades oil, so it is arguably an oil trading business, and its loyalty program makes it closer to Marriott, Hilton and Hyatt, which own no property and sell points.
14. Banks as Tech Stocks
Asked what they are bullish on that had not come up, Lee went to financial services. Everything that happens in the real economy has to be represented on a ledger, and the industry knows it is not equipped for robots making micropayments with no clear instruction source.
Much of that plumbing, he said, gets built on blockchains, citing Robinhood's view that the whole system settles that way and BlackRock and JPMorgan treating blockchain as the future. His conclusion is a re-rating: the best financial companies become technology stocks and should be valued that way now.
The visibility argument
Today on January 1, JPMorgan already knows like 70% of its earnings without even opening a branch.
Tom Lee
Few companies have that, he said, naming Costco as one that does and trades at 50 times. He drew the further conclusion that crypto is a large winner, because that is where the activity settles.
Ives extended it to utilities and energy, and argued the boom is still underappreciated beyond the handful of large technology names, on jobs as much as on spending: he expects more jobs created than taken away, particularly for young engineers.
Asked about the revival of the hardware names of his youth, with Cisco and Dell market caps multiplying, Lee said the lesson is about the durability of a moat.
And where index returns actually come from
30% of the return come from in the S&P comes from companies that didn't exist 10 years ago
Tom Lee
The majority, he said, still comes from value stocks, which produces his conclusion about the index.
Most of the next move is already owned
If someone doesn't think you can get to S&P 16,000, 80% of what'll get us there is already in your portfolio.
Tom Lee
15. Tickers and an IPO
The hosts ran through individual names at the end.
On Apple, after two years of being told the company had no AI strategy, the stock sits closest to a record high among the largest technology names, and the absence of a data-center budget now looks like an advantage.
The distribution argument
I mean I think 20% of the world's going to access AI through an Apple device.
Dan Ives
He compared it to the years when the market gave Apple no credit for services and the app store, and said the new chief executive inherits an AI-enabled device cycle from Tim Cook.
On Palantir, which he said has moved from elementary school toward $200, Ives argued the market underappreciates how advanced the technology is and what it has done to the enterprise sales cycle.
His price expectation
This is a stock that could appreciate four 5x from here over the next 3 to four years given the cash flow.
Dan Ives
He called it the next trillion-dollar name, on free cash flow rather than on any revenue multiple.
On Anthropic, which the hosts expect to come public in October, Lee was asked whether it qualifies for his fund.
Not a Granny Shot
Anthropic is going to be tough to qualify as a granny shot.
Tom Lee
His framing of the frontier labs is that they are all racing because each believes only one wins, the way search resolved to Google, and that an investor probably has to own more than one; if it is worth five trillion, the same argument makes it worth one. The answer from the table was that Anthropic becomes an enterprise company rather than a model company, which is why it is building an enterprise sales force.
And the stage of the cycle
This is a 1997 moment not a 1999 2000 moment.
Dan Ives
Bonus Insights
This was a live show, and the guests were introduced to an audience rather than to a microphone. Fundstrat marked its twelfth anniversary on the day of the recording. Ives co-founded Yorkville Ives and Co. with Yorkville Securities in July, combining investment banking, equity research, institutional trading and principal investment with a focus on AI, after eight years running technology research at Wedbush.
The robot joke of the session was a coffee-making machine at a booth on the conference floor, which Lee used as his example of a working use case before one of the hosts pointed out it might also be the thing that kills them.
The hosts closed by announcing a December event in New York, and asked the audience to subscribe to the show's newsletter for first access to tickets.
The bottom line from both guests is that a supply-constrained chip market, earnings rising faster than prices and a Fed that is tightening into relief rather than into a bubble add up to the early part of a technology cycle rather than the late part, with the political fight over data centers as the risk they rate above valuation.
Products, Companies & Tools Mentioned
Nvidia (The single chip Ives says is fueling the build-out; Lee's point is that it trades at 16 times against 50 for Costco)
Anthropic and OpenAI (The two leaders in models, far enough ahead that Ives says nothing but regulation closes the gap; the hosts expect an Anthropic listing in October)
Palantir (Ives's four-to-five-times call over three to four years, on free cash flow, and the source of the sovereign-AI framing)
Apple (Closest of the largest technology names to a record high; Ives expects 20% of the world to reach AI through an Apple device)
Tesla and SpaceX (Ives puts the odds of a merger by the end of next year above 80%, and calls the combination the largest company in the world)
CrowdStrike, Fortinet and Palo Alto Networks (All up more than 100% on the year, after the market decided a frontier model had ended the industry)
Micron (Where Ives says the memory question is visible in the share price)
Salesforce and ServiceNow (The software names the apocalypse narrative was supposed to destroy, with nothing fundamental having happened)
Adobe and Intuit (The two Ives allows may have genuine structural questions from AI)
JPMorgan and BlackRock (Lee's examples of financial firms treating blockchain as the future; JPMorgan's earnings visibility is his re-rating argument)
Robinhood (Which he says expects the entire system to settle on blockchains)
Huawei (The Chinese comparison: behind even a third-tier US chip by about a year, on his account)
Cisco and Dell (The revived hardware names, used to argue a moat lasts longer than a 12-month horizon)
Costco and Walmart (The multiples Lee holds up against Nvidia's)
Delta (The classification problem: an airline that refines and trades oil and sells loyalty points)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: