Aug 27, 2026 · 30 min
Listen from 9:15
Listen at 9:15
NVIDIA’s surge tests the economics of AI infrastructure
NVIDIA Earnings Blow Everyone Away
NVIDIA’s results are forcing investors to reassess semiconductor bottlenecks, software pricing, and whether AI spending ultimately shifts back toward local computing.
- 1NVIDIA’s guidance underscores extraordinary AI infrastructure demand while exposing supply constraints, memory costs, and margin pressures.
- 2CrowdStrike’s growth and Mythos strategy raise a broader question about AI-driven cybersecurity valuations and usage-based software pricing.
- 3Local AI could challenge centralized data centers as falling hardware costs reshape where computing happens.
Don't miss
The listener question about local versus centralized AI crystallizes the episode’s biggest uncertainty: whether data-center investment eventually gives way to more on-device and in-house computing.
The brief
John Quast joins the hosts to examine NVIDIA’s unexpectedly strong earnings and guidance, asking whether extraordinary AI demand can overcome supply constraints, memory costs, and the infrastructure sector’s high expectations.
The conversation widens from NVIDIA to the companies building around the AI boom, including Google, Hewlett Packard Enterprise, Dell Technologies, and Credo Technology, as semiconductor bottlenecks shape who benefits.
CrowdStrike’s blowout quarter becomes a test case for AI-era software: accelerating recurring revenue and its Mythos initiative support the growth story, while valuation and stock-based compensation complicate the investment case.
The hosts then consider whether AI will replace subscription pricing with usage-based models, a shift that could change how software companies measure growth and monetize their products.
A listener question supplies the episode’s central tension: if local and on-device AI become capable enough, today’s massive data-center buildout could give way to more computing in homes, offices, and private servers.
The closing argument is less that cloud computing disappears than that it may come full circle, with centralized systems and local hardware dividing workloads by economics and use case.
What was said on this episode
22 statements · 13 positive · 6 negative · 1 mixed · 2 neutral
NVIDIA trades at approximately 25 times forward earnings despite 106% growth.
“companies that grow revenue at 106% year over year. generally don't trade for 25 times forward earnings like NVIDIA does.”
Listen at 2:35
Higher memory costs may compress NVIDIA’s margins.
“There are margin compression concerns due to memory costs”
Listen at 3:28
Higher memory costs are problematic for HP.
“I think that that becomes a little bit problematic for a company such as HP.”
Listen at 5:35
HP is poorly positioned amid rising memory costs and shortages.
“I would say that's a tough place to be if you're HP.”
Listen at 6:38
Credo stock could perform well if revenue reaches $5 billion within several years.
“I believe this stock can be a good performer, let's say, if it reaches $5 billion in revenue within the next several years.”
Listen at 7:35
Onto Innovation and similar data-center suppliers will gain pricing power.
“companies like this are going to have Not only a lot to do, but a lot of pricing power.”
Listen at 8:45
Memory shortages could raise Garmin prices or compress margins.
“it could raise prices on Garmin devices or result in margin compression”
Listen at 9:15
AI is increasing cybersecurity demand faster than expected.
“AI is clearly fueling demand for cybersecurity at a much stronger rate than experts thought.”
Listen at 13:30
CrowdStrike’s flex billing may explain Mythos-related ARR growth.
“So maybe this is more of a billing thing that Mythos is showing us why this is such a good way to bill its customers”
Listen at 16:23
CrowdStrike’s valuation should be a lower-priority concern for investors.
“If ever there was a company to not ignore the valuation entirely, but to push it way down your concerns list, I think CrowdStrike is one of them”
Listen at 19:03
Rubrik trades at approximately 14 times sales.
“Rubrik, RBRK. This is a cybersecurity company, but not a direct competitor to CrowdStrike trades at only 14 times sales.”
Listen at 19:41
Stock valuation always matters at least somewhat.
“valuation always matters at least a little bit.”
Listen at 20:13
Investors should not initiate a CrowdStrike position today.
“I wouldn't start a new position here today”
Listen at 21:40
Investors may dollar-cost average into CrowdStrike.
“it's definitely a great case for dollar cost averaging into a position.”
Listen at 21:45
Jon recently bought Zscaler shares.
“I recently bought Zscaler.”
Listen at 21:51
Local AI is not a threat to the cloud AI industry.
“I don't think that local AI is a threat to the cloud AI industry.”
Listen at 26:11
Individuals will increasingly want to run local AI.
“more and more people, individuals are going to want to run some local AI.”
Listen at 26:21
Hospitals, schools, and government agencies will increasingly use local AI.
“hospitals, schools, government agencies, they're all going to want to run more and more local AI”
Listen at 26:48
Cloud AI will remain necessary for some future innovations and tasks.
“I still think there's going to be innovation happening and things that you're going to need a cloud AI for.”
Listen at 27:19
AI data centers will continue growing due to key advantages.
“data centers have a few key advantages that are gonna help them continue to grow.”
Listen at 28:00
Local AI could strongly benefit cybersecurity.
“it could be one of the best things that ever happened to cybersecurity.”
Listen at 28:39
Local AI could benefit CrowdStrike and Zscaler.
“definitely could help companies like CrowdStrike and Zscaler going forward”
Listen at 28:54
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.