NVIDIA earnings beat, Anthropic IPO, and AI cost deflation trends discussed
Summary
Andy Zamphotis covers NVIDIA's Q3 earnings, noting record revenue and raised guidance despite CEO Jensen Huang's departure, with interim CEO Colette Kress emphasizing selective financing. The episode details Anthropic's $83 IPO price and $1.5 trillion valuation, highlighting $10 billion investments from Alphabet and NVIDIA. Google's new Gemini Neon model is presented as a driver of AI cost deflation, priced 25% below competitors while maintaining near-frontier performance. Palo Alto Networks' acquisition of Applied Compute is analyzed as a move toward vertical integration and private AI infrastructure for enterprise clients.
Key takeaways
- 1 Anthropic's IPO raised over $80 billion, with Alphabet and NVIDIA each committing $10 billion at the IPO price in a $26 billion private placement.
Andy Zamphotis · 7:49
- 2 actually own a pretty large stake from an earlier private round. And if you count for the converts that Amazon has, I think their stake is over 20.
Andy Zamphotis · 8:57
Episode sections
- 0:23 – 7:40
NVIDIA Earnings and Leadership Transition
Andy Zamphotis discusses NVIDIA's Q3 earnings, highlighting record revenue and raised guidance despite CEO Jensen Huang's resignation. The interim CEO, Colette Kress, emphasized selective customer financing and raised fiscal 2027 growth estimates to 75%.
- 7:40 – 10:35
Anthropic IPO and Big Tech Investment
Anthropic priced its IPO at $83, raising over $80 billion and valuing the company at $1.5 trillion. Alphabet and NVIDIA each invested $10 billion, giving Alphabet approximately 15% ownership and highlighting the circular financing dynamics in the AI sector.
- 10:35 – 13:31
Google's Gemini Neon and AI Cost Deflation
Google released Gemini Neon, a model priced 25% below competitors like Claude Opus 5.5 while delivering near-frontier intelligence. The speaker notes that cheaper inference costs could drive massive usage growth, benefiting infrastructure providers even if model pricing power declines.
- 13:31 – 15:55
Regulatory Risks and Open Source AI
The discussion covers potential US regulations targeting Chinese AI models and the risk of broader bans on open-source models. Palo Alto Networks' $4.5 billion acquisition of Applied Compute is cited as evidence that major tech firms do not expect a ban on open weights.
- 15:55 – 24:23
Palo Alto Networks' Vertical Integration
Palo Alto reported strong fiscal Q1 results and announced the acquisition of Applied Compute to build its own AI infrastructure. The company plans to invest up to $3 billion in capex through fiscal 2030 to serve enterprise clients with private, on-premise AI models.
Assets discussed
| Asset | Kind | Stance | Context |
|---|---|---|---|
| NVIDIANVDA | equity | neutral | this week we had NVIDIA Report Earnings delivering another extraordinary course. quarter. |
| AmazonAMZN | equity | neutral | Amazon is also another indirect. way. |
| AlphabetGOOGL | equity | neutral | which brings me to the next story, which is Google. |
| MicrosoftMSFT | equity | neutral | And you know, Microsoft and Meta have also backed an industry letter opposing broad limits on open models. |
| Meta PlatformsMETA | equity | neutral | And you know, Microsoft and Meta have also backed an industry letter opposing broad limits on open models. |
“Stance” is how the speakers talked about the asset, not a recommendation.
Claims worth verifying
CEO Jensen Huang resigned after reports about his personal conduct and stated he would be honored to return if the board wanted him.
Unidentified speaker (UNKNOWN) ·1:38
How to verify: SEC Form 8-K filing for NVIDIA Corporation (NVDA) regarding executive resignations; press releases from NVIDIA and news outlets covering the event. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceThe board does not intend to reinstate Jensen Huang.
Unidentified speaker (UNKNOWN) ·1:51
How to verify: NVIDIA Q3 FY2026 Earnings Call transcript; SEC filings or press releases from NVIDIA. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceNVIDIA's October quarterly revenue was $115 billion, up over 100% from the prior year.
Andy Zamphotis ·2:53
How to verify: NVIDIA Q3 FY2026 Earnings Press Release and 10-Q filing with the SEC. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic priced its IPO at $83 per share, raising over $80 billion.
Unidentified speaker (UNKNOWN) ·7:48
How to verify: SEC S-1 filing for Anthropic AI; press releases from Anthropic and financial news outlets. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAlphabet and NVIDIA each committed $10 billion to a $26 billion private placement at Anthropic's IPO price.
Andy Zamphotis ·8:31
How to verify: SEC filings for Alphabet (GOOGL) and NVIDIA (NVDA) disclosing investments; press releases from Anthropic. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceGoogle's Gemini Neon model costs about $3 per task, compared to $10 per task for Claude Opus 5.5.
Unidentified speaker (UNKNOWN) ·11:02
How to verify: Google Cloud pricing pages for Gemini; Anthropic pricing pages for Claude; third-party benchmarks like Artificial Analysis. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Palo Alto Networks agreed to acquire Applied Compute in an all-stock deal worth $4.5 billion.
Unidentified speaker (UNKNOWN) ·18:59
How to verify: SEC S-4 filing for Palo Alto Networks (PANW) regarding the acquisition; press releases from Palo Alto Networks. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourcePalo Alto Networks targets cumulative capex under $3 billion through fiscal 2030 for AI infrastructure.
Unidentified speaker (UNKNOWN) ·20:25
How to verify: Palo Alto Networks Fiscal Q1 FY2026 Earnings Call transcript; investor presentations. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Do your own research
- Verify NVIDIA's Q3 2025 earnings report and press release for revenue figures and guidance updates.
- Review Anthropic's S-1 filing for IPO pricing, investor commitments, and ownership stakes.
- Check Google's official announcement for Gemini Neon pricing and performance benchmarks.
- Examine Palo Alto Networks' Q1 fiscal 2026 earnings release for Applied Compute acquisition details.
- Assess regulatory filings for any US government actions regarding Chinese AI models or open-source restrictions.
- Compare inference pricing data from major AI providers (OpenAI, Anthropic, Google) to validate cost deflation claims.
- Review NVIDIA's investor presentation for details on customer financing structures and Vera Rubin shipment volumes.
- Analyze Palo Alto Networks' capex guidance and infrastructure plans for fiscal 2030.
Full transcriptexpand
0:04Unidentified speaker (UNKNOWN): Welcome
0:04Andy Zamphotis: to the AMI investment perspective for November the 20th. I'm Andy Zamphotis, portfolio manager and tech analyst.
0:12Unidentified speaker (UNKNOWN): This
0:12Andy Zamphotis: week may have told us a lot about where the AI trade is going more than any week that I can remember, at least in recent history. And it's been a very rocky few weeks, so there's a lot to discuss. Um, So let's dive right
0:22Unidentified speaker (UNKNOWN): in. So
0:23Andy Zamphotis: this week we had NVIDIA Report Earnings delivering another extraordinary course. quarter. Um, but investors continue to ask what the company looks like without CEO Jenzin Wang.
0:35Unidentified speaker (UNKNOWN): Uh
0:35Andy Zamphotis: Anthropic completed its IPO, the largest IPO on record, at a valuation that assumes some you know fairly good future success. And um at the same time, Google released a model delivering near. frontier intelligence at far less per task in terms of cost than some of its competitors.
0:56Unidentified speaker (UNKNOWN): And then
0:56Andy Zamphotis: there's Palo Alto Networks, a cybersecurity company that we've talked about on this podcast quite a bit.
1:02Unidentified speaker (UNKNOWN): They
1:03Andy Zamphotis: agreed to buy an AI model business and are starting to build its own computing capacity.
1:08Unidentified speaker (UNKNOWN): And
1:08Andy Zamphotis: so these aren't necessarily all unrelated stories. They're um about the same question as to, you know. which is as intelligence gets more useful and cheaper and more widely available.
1:20Unidentified speaker (UNKNOWN): Who
1:20Andy Zamphotis: ultimately earns the returns and who has to fund all this?
1:24Unidentified speaker (UNKNOWN): Um,
1:24Andy Zamphotis: so with those questions, let me start with NVIDIA, which I think was the big story for the week. And as I'm sure many are now aware, CEO Jensen Wang resigned last week after reports about his personal conduct.
1:38Unidentified speaker (UNKNOWN): Um
1:38Andy Zamphotis: On Tuesday, Tuesday morning, he publicly said, I think it was his first public statement since his resignation, but he said he'd be honored to come back if the board wanted
1:47Unidentified speaker (UNKNOWN): him.
1:48Andy Zamphotis: And then a few hours later, on the earnings call,
1:51Unidentified speaker (UNKNOWN): uh
1:51Andy Zamphotis: the new board chairman Harvey Jones opened by saying that the board does not intend to reinstate
1:56Unidentified speaker (UNKNOWN): him. So
1:57Andy Zamphotis: there's a lot more going on here. Um, there's gonna be some congressional inquiries, uh, reporterly reportedly maybe. an SEC probe into some accounting, some tighter export rules.
2:07Unidentified speaker (UNKNOWN): Some
2:08Andy Zamphotis: of this is just, you know, uh theory or hypothesis as to what might happen, but um Jensen is scheduled to
2:15Unidentified speaker (UNKNOWN): uh
2:16Andy Zamphotis: testify in front of Congress in early December.
2:19Unidentified speaker (UNKNOWN): Um
2:20Andy Zamphotis: but I don't want to spend too much time on the scandal itself. I think the investment question is whether NVIDIA keeps its operating and strategic edge. throughout this transition, regardless of what happens.
2:30Unidentified speaker (UNKNOWN): So
2:30Andy Zamphotis: Colette Kress, the 13-year CFO and now interim CEO, I think passed her first test. She was steady and specific.
2:41Unidentified speaker (UNKNOWN): You
2:41Andy Zamphotis: know, maybe sounded a little bit nervous at times, which is understandable, but
2:45Unidentified speaker (UNKNOWN): and
2:45Andy Zamphotis: she didn't really try to imitate Jensen. And the numbers were extraordinary. The October quarterly revenue
2:52Unidentified speaker (UNKNOWN): was
2:53Andy Zamphotis: 115 billion,
2:54Unidentified speaker (UNKNOWN): up
2:55Andy Zamphotis: over 100% from the year prior. Vera Rubin did about 20 billion its first quarter of shipments. And the fourth quarter guide was 130 billion, which is ahead of consensus. And they actually raised fiscal 2027 growth from 70% to 75%. And this is despite all the noise surrounding potential dealing. dealings or shipments to China via Singapore, which they somewhat broke
3:19Unidentified speaker (UNKNOWN): out
3:21Andy Zamphotis: and said, you know, a lot of this guidance didn't assume compute to China.
3:26Unidentified speaker (UNKNOWN): Yeah,
3:26Andy Zamphotis: I mean, one would think if there was a lot of controlled exports inflating numbers, um, somewhat, you know, dodgy, the guide would be in jeopardy and it wasn't um it was raised, and there were some leading up into the earnings report that the company might even decide to pull guidance for some of those reasons. So they didn't, they didn't pull it, they decided to raise it.
3:57Unidentified speaker (UNKNOWN): Um
3:57Andy Zamphotis: They don't necessarily prove that NVIDIA can execute as well without its founder and champion. Um But the guide is certainly taking to consideration, you know, after he's gone or after he left. And it's at least for now it's saying that the financial engine is certainly not stalling.
4:12Unidentified speaker (UNKNOWN): Um
4:13Andy Zamphotis: but what I think what caught my attention is certainly some folks on the street were looking into was the financing discussion. So before earnings,
4:20Unidentified speaker (UNKNOWN): there
4:21Andy Zamphotis: was There's an article in the information that reported that Cress, as when she was CFO, pushed back on a proposed $250 billion credit backstop for OpenAI. And you know, that made people wonder whether she'd pull back on what some have been calling like the central bank of AI or the bank of NVIDIA. And these are the various investments and guarantees that help customers buy NVIDIA's products.
4:43Unidentified speaker (UNKNOWN): So
4:44Andy Zamphotis: the thinking there is without Jensen in the ch in the seat,
4:48Unidentified speaker (UNKNOWN): you know,
4:49Andy Zamphotis: what does that mean for all these deals? Um, and so on the call, what you know, what I heard was perhaps more selectivity, not necessarily reversal.
4:57Unidentified speaker (UNKNOWN): Um,
4:57Andy Zamphotis: again, we don't know who the permanent CEO will be and if he or she ends
5:02Unidentified speaker (UNKNOWN): up
5:02Andy Zamphotis: changing those
5:04Unidentified speaker (UNKNOWN): um
5:04Andy Zamphotis: directives further, but
5:06Unidentified speaker (UNKNOWN): at
5:06Andy Zamphotis: least from what Colette Krest said, that there's there seems to be a preference for investment grade customers and prepayments and guarantees tied to funded projects, not speculative deals. And when asked whether whether anything had changed,
5:20Unidentified speaker (UNKNOWN): um,
5:21Andy Zamphotis: she had said, you know, this is a quote, is it new? No. Did we change anything? No. So, you know, cleanup of customer financing
5:29Unidentified speaker (UNKNOWN): is
5:29Andy Zamphotis: just a theory at this point, not a fact. And at least for now, it's kind of mixed. So, you know, guaranteed exposure was up in the quarter. A lot of that was tied to memory commitments. Um, but agreements with neo clouds seem to have dipped, which is why you know how these things go in the next couple of quarters will be important, um, especially
5:52Unidentified speaker (UNKNOWN): if
5:52Andy Zamphotis: there's a new CEO. Um, because if customer financing becomes a constraint, you know, that's going to matter more than chip supply.
6:01Unidentified speaker (UNKNOWN): Um, and
6:02Andy Zamphotis: then near you know, so the the they tied the next year's outlook, which is the 75% growth figure I mentioned, to several conditions. One was data centers actually getting built, demand holding up, and customers keeping access to capital. So those are three, you know, I think pretty important conditions.
6:18Unidentified speaker (UNKNOWN): You
6:18Andy Zamphotis: know, despite all this, the stock continues to you know somewhat someone meander to the downside. Um, you know, after they reported earnings, it was up a little bit, but gave it all back. And then some. And I think the reason is that while NVIDIA's current business is clearly strong, the chips remain the best in the business,
6:37Unidentified speaker (UNKNOWN): at
6:38Andy Zamphotis: least for now.
6:39Unidentified speaker (UNKNOWN): Um,
6:39Andy Zamphotis: you know, it's leadership model, the customer financing exposure, and the policy risk matters, whether, you know, how and what we can ship or to China. Though those those questions are less settled. And you know, the call told
6:55Unidentified speaker (UNKNOWN): us
6:56Andy Zamphotis: somewhat of some things about continuity, but we'll learn a little bit more in the next few weeks after we get this testimony from Jensen to Congress as to whether or not something perhaps untoward was happening. And then
7:11Unidentified speaker (UNKNOWN): the
7:11Andy Zamphotis: company didn't do itself any favors by postponing the 10 10Q, key which I think to some was understandable given all that's gone on when the CEO has resigned abruptly. But,
7:23Unidentified speaker (UNKNOWN): you
7:23Andy Zamphotis: know, seeing if there's any caveats or new risk factors being added, I think people want to see that as well. But the market's going to need more because to truly believe those growth figures of 75%, I think investors have to get comfortable that the dust is settled.
7:39Unidentified speaker (UNKNOWN): So
7:40Andy Zamphotis: the other big headline this week was anthropic. They priced its IPO at $83 a share, which is above the range. Original target was like in the mid to low 70s.
7:48Unidentified speaker (UNKNOWN): They
7:49Andy Zamphotis: raised over $80 billion, largest IPO.
7:52Unidentified speaker (UNKNOWN): And at
7:52Andy Zamphotis: that IPO price, it was worth about $1.5 trillion on a fully diluted basis. Stock opened yesterday, above $110, I think it was around $111. which is about 30, 30-ish percent from the IPO price,
8:06Unidentified speaker (UNKNOWN): which
8:06Andy Zamphotis: put the market cap at about 2 trillion. And I think the stock went as high as 115, but ended up closing under 100. Today, as I record this, it's down further, closer to the offering price, around $85. But the whole age AI infrastructure trade is languishing today and is down for a whole host of reasons. Some of it's macro. like rates
8:27Unidentified speaker (UNKNOWN): and
8:28Andy Zamphotis: other things. But um
8:30Unidentified speaker (UNKNOWN): but you
8:31Andy Zamphotis: know, the short term trading aside, what I found most interesting is that NVIDIA and Alphabet committed 10 billion each to in a 26 billion private placement at the IPO price. So
8:41Unidentified speaker (UNKNOWN): what
8:41Andy Zamphotis: ends up happening here is Alphabet will own something like 15% of Anthropic, which is a nice indirect way to get exposure.
8:48Unidentified speaker (UNKNOWN): Um,
8:48Andy Zamphotis: Amazon is also another indirect. way. They didn't, I don't believe they participated in this private placement at the IPO, but
8:56Unidentified speaker (UNKNOWN): they
8:57Andy Zamphotis: actually own a pretty large stake from an earlier private round. And if you count for the converts that Amazon has, I think their stake is over 20. So you have big tech somewhere in the order of like owning a third of the company. And Um yeah, you know again, all these companies are helping to support a very large customer that buys chips and buys compute. So, you know, while it certainly helps their business, it doesn't help the whole circular financing argument that some of the
9:20Unidentified speaker (UNKNOWN): some
9:21Andy Zamphotis: of the bearish folks have been making. But that's that's a debate I'll save for another day.
9:26Unidentified speaker (UNKNOWN): Yeah,
9:26Andy Zamphotis: and some thought this IPO, which you know, by all accounts was a s was a success, would lift the whole AI trade. It might be a clearing event. It didn't really help yesterday or today for that matter, but perhaps over time, you know, that we have a well funded well-funded customer. customer that can now pay its bills, which, you know, was a big concern not too long ago. The question was, you know, Anthropic was making all these commitments and didn't have the cash.
9:48Unidentified speaker (UNKNOWN): I think
9:48Andy Zamphotis: they have a war chest of well over $200 billion at this
9:51Unidentified speaker (UNKNOWN): point.
9:52Andy Zamphotis: Um, but that doesn't seem to be helping the broader, the broader trade. And it certainly hasn't helped software companies, which now face a very well-funded AI threat to their core business. So we've seen a bunch of this set. the SaaS players like a Salesforce and other companies go down.
10:07Unidentified speaker (UNKNOWN): But
10:07Andy Zamphotis: above all, I think having anthropic public is now a good way to track the economics of this whole AI trade, the growth, which has been another which has been the source of another debate as to whether or not this stuff can be profitable or not. You know, the financials have been impressive, but you know, at a trillion plus valuation, we need more than just solid revenue groups growth. We have to see if pricing power can hold up as competition gets more intense, which brings me to the next story, which is Google.
10:35Unidentified speaker (UNKNOWN): So
10:35Andy Zamphotis: Google made a splash this week. They released their new Gemini for Neon,
10:40Unidentified speaker (UNKNOWN): which
10:40Andy Zamphotis: is part of what it calls a periodic table of models. They had carbon and hydrogen.
10:44Unidentified speaker (UNKNOWN): So
10:45Andy Zamphotis: these are models that are tuned for very different jobs instead of one model for everything.
10:49Unidentified speaker (UNKNOWN): So
10:49Andy Zamphotis: hydrogen is the cheap, fast work. workhorse, carbon is the heavy reasoner, and then neon, which is what came out this week, sits in the middle. It's priced about 25% below carbon and um
11:02Unidentified speaker (UNKNOWN): and
11:02Andy Zamphotis: Claude Opus 5.5, which was considered like the sort of trendsetter, at least. Um, and the reviews have been strong. So on the artificial analysis intelligence index, which is a you know widely tracked. benchmark, so neon is you know sits high, setting scores at about 60, at about a little over $3 per task. So just by comparison, Claude Opus 5.5, again, which is considered like one of the best models out there,
11:28Unidentified speaker (UNKNOWN): um,
11:28Andy Zamphotis: at max effort, scores about 59,
11:31Unidentified speaker (UNKNOWN): um,
11:32Andy Zamphotis: which obviously very close, but at a cost of about $10 per task. So you know, we're talking huge discount per task. And you know, the settings differ, so treat that more as directional. And again, OPAS still beats NEON and some benchmarks, but the directionality is pretty clear. So you have near near frontier intelligence getting much cheaper and faster. And I think that's the tension long-term for an anthropic shareholder in terms of what they do. So you're paying for an elite model provider, just as these hyperscalers are making near frontier capabilities. capability available at a much lower cost per task, which they can afford to do given their financial resources. And they obviously own their own compute.
12:11Unidentified speaker (UNKNOWN): But
12:12Andy Zamphotis: Anthropic's coding strength and enterprise relationships at this point might protect it longer term. But I guess the proof will show up over time in terms of its pricing and margins. So another good reason to have them be public so we can all see this white in the in the open. But there's a second side to these falling prices, and it may be, I think, the more important point this week. So, you know, cheaper inference unlocks a lot of new use cases. So we get more agents. I talked about Muse and, you know, other agents on this podcast,
12:43Unidentified speaker (UNKNOWN): uh,
12:43Andy Zamphotis: longer workflows and far more tokens before doing anything for an extended period of time just costs cost too much money. So if the cost per task you know falls by half, but usage rises like five or 10x, you know, total spending rises. So separate model pricing pressure from infrastructure demand, I think, is the key point there. Um, and Google again has its models, it has its cloud capacity, it certainly has distribution,
13:09Unidentified speaker (UNKNOWN): and
13:09Andy Zamphotis: now it has a stake in an outside model company. So it's it's hedged across the entire stack, but I wouldn't say it's it's it's mune immune, like you know, if prices fall too fast, and certainly if they if it falls fast and usage is growing, then you know, the
13:24Unidentified speaker (UNKNOWN): then
13:24Andy Zamphotis: both it's stake and anthropic and the return on its own, um AI catbacks can feel it. So
13:31Unidentified speaker (UNKNOWN): but
13:31Andy Zamphotis: I think overall, this is why as a firm, we preferred only the hyperscalers um given that they have these advantages. And then one more wrinkle this week, which is more political. So the White House replaced its national cyber director, this guy, Sean Cairncross, I think is how you pronounce his last name. And there was, you know, this is again amid a week of major outages that have become very public. And then separately, the administration is weighing a rule that would make US companies have to disclose when they use Chinese. AI models.
14:04Unidentified speaker (UNKNOWN): Um, and
14:05Andy Zamphotis: that there's been talk all year of going further, like all the way to a ban potentially of these Chinese models.
14:11Unidentified speaker (UNKNOWN): Um
14:12Andy Zamphotis: and you know, that's fueled chatter about regulatory capture. Is this the first step toward banning open source AI, which uh effectively will then hand the market to anthropic, um, Google, and and open OpenAI. AI. Um so you know.
14:27Unidentified speaker (UNKNOWN): that
14:27Andy Zamphotis: may happen, but I don't think so, right? At least not yet. So everything on the table targets models by country of origin, not necessarily open weights in general as a category. So NVIDIA has an open weight model. And you know, Microsoft and Meta have also backed an industry letter opposing broad limits on open models. Now, Anthrop Anthropic didn't sign that letter, which is part of partly why this red capture talk is. exists in the first place. But it, you know, but they've said they've never advocated a ban.
14:57Unidentified speaker (UNKNOWN): And
14:58Andy Zamphotis: this summer, there, you know, there were some commerce export controls which took Anthropic's some of their newest models offline for more than two weeks. Um, I think we if you guys recall that when Fable first came
15:09Unidentified speaker (UNKNOWN): out. Um, you
15:10Andy Zamphotis: know, but a crackdown on these Chinese models, and again, this is these are like DeepSeek and Kemi. Kimi and other other models that
15:16Unidentified speaker (UNKNOWN): are
15:17Andy Zamphotis: essentially distilled versions of some of the Frontier models, but they can offer them cheaply. And it matters for pricing. So eight of the 10 most used models on OpenRouter this month are of Chinese origin. And they win largely because they're cheap. So if you take them out and then you get pricing power improving for every US lab, whether it's closed or open.
15:37Unidentified speaker (UNKNOWN): And
15:38Andy Zamphotis: I guess what would change my views on this is that if if rules that restrict open weights by capability rather than by country, so you know, what we don't want to see is a pre-release review for American open models or you know, the Commerce Department putting Chinese lab on its entity list. And um, and this matters. And we just, you know, there's one company that just put four and a half billion behind the idea that open models aren't going anywhere. And that's Palo Alto, which is um where I want to go next. So Palo Alto reported very strong fiscal Q1. Stock was up a lot after hours on those results. Revenue was up near 40%, next gen security up 65%. They raised their outlook. And even on an organic basis, growth was in the mid to high 20s.
16:22Unidentified speaker (UNKNOWN): Um,
16:23Andy Zamphotis: so, you know, and then CEO Nikesh Aurora's. you know, selling these cybersecurity products and services into an environment that feels chaotic. So on Monday, um, attackers hijacked a website testing tool
16:37Unidentified speaker (UNKNOWN): uh
16:37Andy Zamphotis: which pushed malicious code onto the sites of about 170 organizations like Zoom and Salesforce, CNN, some banks.
16:45Unidentified speaker (UNKNOWN): Um,
16:46Andy Zamphotis: and Nikesh had posted on X saying that the events of this week. so far are not a coincidence. AI is making it much easier for hackers to attack multiple customers.
16:57Unidentified speaker (UNKNOWN): And
16:57Andy Zamphotis: then to be fair, not every outage is an attack. What we saw happen with the flights in the FAA, which is all blamed on hardware upgrades.
17:03Unidentified speaker (UNKNOWN): You know,
17:03Andy Zamphotis: Microsoft said last week's their outlook outage wasn't a cyber attack. And, you know, on the call, Nick Nikesh Cash at Palo Alto himself blames some of the outages on the flood of patches and configurations. configuration changes that are taking place all at the same time. But you know, the numbers are are striking. Um, you know, so
17:21Unidentified speaker (UNKNOWN): he
17:22Andy Zamphotis: said that severe v vulnerabilities are running at like seven times last year's average. And agentic traffic on Palo Alto's network is up about like nine times. And It so um it's it's big. These networks just are being trying to be patched. like really fast and quickly because of the all the different attack surface areas that AI enables.
17:44Unidentified speaker (UNKNOWN): Um
17:45Andy Zamphotis: and it's the the hardware just like not not ready for it just yet.
17:51Unidentified speaker (UNKNOWN): Um
17:52Andy Zamphotis: and
17:53Unidentified speaker (UNKNOWN): so
17:53Andy Zamphotis: but this is also why I think Palo Alto is doing well because they have these platform deals, which you know, again, they had a record quarter of
18:01Unidentified speaker (UNKNOWN): them.
18:01Andy Zamphotis: And so
18:03Unidentified speaker (UNKNOWN): I
18:03Andy Zamphotis: think why this favors like the platform players, and platform by platform, I mean they can offer you coverage for a variety of different areas, whether it's in-house, online, in the cloud, uh, on device.
18:17Unidentified speaker (UNKNOWN): So
18:17Andy Zamphotis: the more complex the threat environment gets with AI, the bigger incentive it is to consolidate security vendors. You don't
18:23Unidentified speaker (UNKNOWN): want
18:24Andy Zamphotis: 10 different companies securing. your system without any sort of coordination. And that's the pitch. And so Nicash and Polito were very early with that platform approach. And as some may recall, this is, I think, about a couple years ago, the stock took a massive hit when they announced it because they had to like slow some revenues for some cut for some renewals in order to get these customers on board. But it's certainly paying off now, which brings you back. me to that four and a half billion dollar bet that I just talked about. So Palo Alto, as part of its earnings call, said that they agreed to acquire a company called Applied Compute, all stock deal.
18:59Unidentified speaker (UNKNOWN): And
19:00Andy Zamphotis: so Applied Compute post-trains and serves open weight models. And they just passed $100 million in annualized revenue, just within about 18 months of the company being founded.
19:10Unidentified speaker (UNKNOWN): And
19:11Andy Zamphotis: Palo Paladin's Alto's argument here is that enterprises have all this proprietary data. they have internal playbooks, and all this like institutional judgment that they can't expose to public models. Like that's that's basically their alpha.
19:22Unidentified speaker (UNKNOWN): And
19:22Andy Zamphotis: enterprises want control over how it's customized, data governance, and cost, of course. So enterprises want to tune their own open models and their own data and run them themselves. So Nikesh said that if this works,
19:38Unidentified speaker (UNKNOWN): you know,
19:38Andy Zamphotis: it could be bigger than what Palo Alto does today. And I don't think he would make this bet if he thought there was going to be a ban on open models anytime soon. Um again, you know, this is a US-based open model company. And
19:51Unidentified speaker (UNKNOWN): I
19:51Andy Zamphotis: think it's useful, it's a useful tell on how one of the more plugged in CEOs and security kind of reads the room in terms of policy risk. And, you know, the issue, I think, for software more broadly is now they're finding that they need to own more of the AI stack, which means owning more of the capital burden.
20:08Unidentified speaker (UNKNOWN): So
20:08Andy Zamphotis: a security company, Powell's not talking about, you know, buying compute and talking about megawatts on these on these earnings calls.
20:15Unidentified speaker (UNKNOWN): Um,
20:15Andy Zamphotis: and they want, you know, about a hundred megawatts of inference over the next several years.
20:20Unidentified speaker (UNKNOWN): They've
20:20Andy Zamphotis: talked about Talking cumulative capex under 3 billion through the fiscal 2030,
20:25Unidentified speaker (UNKNOWN): some
20:26Andy Zamphotis: operating market. margin, you know, operating margin pressure at the peak. You know, management still calls this very capital light. I mean, the logic there is that, you know, 3 billion over four years against a company that's doing about five and a half billion of cash flow to this year alone, pretty small. Um, and they can do this in stages, whether they can either build it themselves or renting from cloud partners in the meantime. Um, so you know, we're seeing the sort
20:51Unidentified speaker (UNKNOWN): of, you know,
20:52Andy Zamphotis: own versus rent in practice. So they're gonna rent while demand is uncertain and then own it once they get more predictability.
20:58Unidentified speaker (UNKNOWN): And
20:58Andy Zamphotis: it it kind of fits the argument we've been making here for a while is that the token makers are getting paid before the token buyers, and now software margins may have to absorb some of these new AI economics. And Palo Alto is making the making the bet that it can become a token maker for its own customers. And its customers, because they want control, they don't want to just hand over all their data and proprietary info to anthropic and OpenAI. They want to do it themselves because they like they like the sort of story behind AI and what it can do. They just don't want to sort of give away the keys to the kingdom. But we'll need to watch these CapEx trends over time. You know, software's main appeal was it being a capitalite recurring revenue business. So why watching what this does to free cash flow and margins over time will be important. And obviously keeping a track of what how how big applied compute gets. Palo's targeting it to be about a $1 billion AR business in a few years. And so if we see other data rich software platforms follow,
21:57Unidentified speaker (UNKNOWN): So you
21:57Andy Zamphotis: know, that's another source of compute demand. And um, but also a reason to think you know the software. margins might take a hit over the next few years.
22:06Unidentified speaker (UNKNOWN): Um
22:06Andy Zamphotis: so to close and just to kind of recap, um, you know, NVIDIA showed the infrastructure engine is still very strong, even though we have questions about leadership,
22:15Unidentified speaker (UNKNOWN): c
22:15Andy Zamphotis: customer financing, and what's happening in DC.
22:19Unidentified speaker (UNKNOWN): Um, you know,
22:20Andy Zamphotis: Anthropic with a pretty pretty good IPO,
22:23Unidentified speaker (UNKNOWN): um,
22:23Andy Zamphotis: showing public investors that you know that they can still get access to a frontier model, albeit at a fairly rich valuation. And then Google showed how we can have cheaper AI and how that can be both a competitive threat but also a demand catalyst. And then Palo Alto in terms of what might be next for software companies.
22:45Unidentified speaker (UNKNOWN): And so
22:45Andy Zamphotis: despite all the volatility we're seeing in some of these names,
22:49Unidentified speaker (UNKNOWN): um,
22:49Andy Zamphotis: you know, for these reasons, I still remain, you know, constructive on AI infrastructure because the demand is still there.
22:56Unidentified speaker (UNKNOWN): The
22:56Andy Zamphotis: open router token usage is there.
22:59Unidentified speaker (UNKNOWN): Um,
23:00Andy Zamphotis: intelligence is getting cheaper, not more expensive. Um,
23:04Unidentified speaker (UNKNOWN): but
23:04Andy Zamphotis: I think the analysis here has to get even more discerning and discriminating than before. Um, before we could just buy anything with AI attached to it and watch the stock go up. And you know, I think what people are realizing is that spending on compute is not the same thing as earning a return on that compute.
23:21Unidentified speaker (UNKNOWN): You know,
23:21Andy Zamphotis: the winners will be the companies that can turn cheaper intelligence into durable revenue, better margins, more free cash flow. And so quality growth, cost of capital will matter even more. Uh And so that's something to watch over these next few quarters.
23:36Unidentified speaker (UNKNOWN): And
23:37Andy Zamphotis: certainly we'll all be paying attention to.
23:39Unidentified speaker (UNKNOWN): Uh
23:39Andy Zamphotis: so that's it for me. I want to thank everyone for listening. Uh we're off next week for the Thanksgiving holiday. So I hope everyone has a uh very good uh holiday and break, and we'll see you on the next episode. Thanks.
23:53Unidentified speaker (speaker_1): AMI is an SEC registered investment advisor. This podcast is provided for general information purposes only, contains no investment recommendations, and should not be construed as individualized tax, legal, financial planning, or investment advice. We may discuss stocks and investment strategies, and please be aware that we might own some of the stocks mentioned, but any mention does not constitute a recommendation to buy or sell the stock. This podcast does not provide personalized financial advice. As individual circumstances vary, you should consult a financial advisor to determine which strategies or investments may be suitable for you.