Behind the video
The UNTHINKABLE is About to Happen to the Stock Market
Overall Thesis
Kevin argues that financing stress at indebted AI infrastructure companies does not signal the end of the broader AI investment boom, and he expects another euphoric market rally before an eventual bubble ends. He favors financially strong hyperscalers, citing sustained infrastructure spending, rising ASML customer deposits, improving AI usefulness, and potentially stabilizing memory costs.
Narratives
METAMeta PlatformsKevin includes Meta among his preferred hyperscaler investments because its financial resources make it less dependent on geopolitical relief. Continued data center spending and custom chip development support his favorable view.
Key Arguments
- Meta can continue funding infrastructure despite elevated borrowing costs.
- Its large data center commitments support the broader chip buildout.
- Stabilizing memory costs could benefit its AI infrastructure economics.
Risks acknowledged
- The broader AI investment cycle could eventually become a bubble.
- Infrastructure and chip availability can constrain deployment.
Predictions (1)
CRWVCoreWeaveKevin views CoreWeave as particularly exposed to expensive debt financing and says neo-clouds are unsuitable investments without an Iran agreement. This is a conditional avoidance and financing-risk thesis rather than an explicit forecast that CoreWeave shares will decline.
Key Arguments
- Rising credit default swaps accompany pressure on bonds and shares.
- Debt-funded expansion makes CoreWeave especially sensitive to high yields.
- Continued hyperscaler spending could sustain the wider boom even if CoreWeave fails.
Risks acknowledged
- An Iran agreement could lower funding costs and allow neo-clouds to participate in the upside.
- Its problems do not establish that the whole AI market will collapse.
Kevin interprets a sharp increase in ASML contract liabilities as evidence that customers are committing to more chipmaking equipment. His commentary supports an optimistic business outlook without establishing a specific forecast for ASML shares.
Key Arguments
- Reported contract liabilities rose from roughly €3.5 billion to €9.3 billion.
- Customer deposits provide an interim indicator of equipment demand.
- Potential chip shortages could encourage further manufacturing capacity investment.
- He compares an approximately 35 forward earnings multiple with expected growth around 24%.
Risks acknowledged
- Deposits are an imperfect substitute for disclosed backlog.
- He is uncertain which institution supplied the stronger-backlog estimate.
- Infrastructure forecasts may face delays.
AMDAdvanced Micro DevicesKevin expects AMD's advanced chips to participate in continued AI infrastructure expansion. He sees stabilizing memory prices and more efficient model deployment as favorable business conditions without separately forecasting AMD investment returns.
Key Arguments
- Hyperscalers continue seeking additional chips.
- Smaller models and efficient memory configurations broaden deployment opportunities.
- Stabilizing memory costs could benefit chip designers.
Risks acknowledged
- Chip production and infrastructure availability can remain bottlenecks.
- The broader investment cycle could eventually become a bubble.
AVGOBroadcomKevin mentions Broadcom as a participant in Google's custom accelerator ecosystem. Its role supports his broader chip-demand thesis, without a separate investment-return prediction.
Key Arguments
- Custom accelerator development broadens the AI chip buildout beyond general-purpose GPUs.
- Continued hyperscaler spending supports the surrounding chip ecosystem.
MSFTMicrosoftKevin explicitly favors Microsoft as a financially strong hyperscaler that remains investable with or without an Iran agreement. He expects its resources to support continued AI infrastructure spending despite stress at smaller providers.
Key Arguments
- Strong finances reduce dependence on lower near-term funding costs.
- Hyperscaler spending can sustain the AI buildout even if neo-clouds struggle.
- Microsoft's withdrawal from an nScale project reflects project and power delays rather than an end to overall AI demand.
Risks acknowledged
- An Iran agreement would improve the environment more quickly.
- Infrastructure timing remains uncertain.
Predictions (1)
AMZNAmazonKevin recommends Amazon alongside other well-funded hyperscalers as an investment that does not require an Iran agreement to work. He expects sustained infrastructure spending from this group to keep the AI expansion going.
Key Arguments
- Financial strength limits dependence on falling financing costs.
- Large infrastructure commitments distinguish hyperscalers from smaller neo-cloud providers.
- More available memory could improve deployment economics.
Risks acknowledged
- Geopolitical relief would accelerate improvement.
- The investment boom could eventually end in overbuilding.
Predictions (1)
ORCLOracleKevin associates Oracle's elevated credit default swaps with financing anxiety and current stock pressure. He nevertheless says its balance sheet has improved and treats its large data center leases as evidence that AI infrastructure demand remains strong.
Key Arguments
- Credit stress can raise the cost of financing expansion.
- Oracle is less rate-sensitive than CoreWeave in his assessment.
- Large Oracle leases helped initiate a wave of industry infrastructure commitments.
Risks acknowledged
- An Iran agreement could alleviate yield-related fears.
- Company-specific financing stress does not establish an industry-wide demand collapse.
NVDANvidiaKevin presents Nvidia as attractively valued relative to expected growth and positioned to benefit from sustained GPU demand. Lower memory requirements per GPU and stabilizing memory prices could improve supply and deployment economics, but he gives no independent explicit share-price forecast.
Key Arguments
- He cites a forward earnings multiple near 23 and expected growth around 35%, producing a PEG near 0.65.
- The infrastructure expansion could lead to another GPU shortage.
- Less memory per GPU could allow more GPUs to be produced.
- Increasingly useful AI applications support demand.
Risks acknowledged
- The broader boom could eventually produce overbuilding.
- The benefits depend partly on technical configurations and memory availability.
GOOGLAlphabetKevin cites Google's data center commitments and custom accelerators as evidence that AI investment extends beyond distressed neo-cloud providers. More available memory could support its infrastructure expansion, but he does not give a distinct Alphabet stock-direction call.
Key Arguments
- Google is among the large buyers committing to data center capacity.
- Custom accelerators contribute to demand for chip manufacturing.
- Stabilizing memory prices could benefit its AI deployments.
Risks acknowledged
- Memory may merely stop rising as quickly rather than become cheaper.
- The expansion could eventually lead to overbuilding.
MRVLMarvell TechnologyKevin identifies Marvell as another company involved in the custom accelerator ecosystem. The mention reinforces his operational optimism about AI chips rather than providing a distinct Marvell stock forecast.
Key Arguments
- Custom accelerators represent another channel for hyperscaler chip investment.
- The host expects infrastructure spending to continue despite neo-cloud financing stress.
Hedges & Caveats
- An Iran agreement is uncertain; renewed conflict could worsen yields and hurt the AI trade in the short term.
- Neo-cloud financing remains vulnerable if funding costs stay elevated.
- Data center completion estimates could slip, and available power does not guarantee available chips.
- SemiAnalysis can have an overly optimistic industry outlook.
- The AI buildout could eventually become an overbuild and a bubble.
- ASML backlog expectations rely partly on deposits and outside estimates.
- Memory prices may merely stop accelerating rather than become cheaper.