Behind the video
They JUST Accidentally Revealed the Truth: The AI Bubble
Overall Thesis
Kevin views SK hynix's earnings commentary as strong evidence that AI infrastructure demand is broadening and becoming more durable, with long-term contracts and custom memory potentially tempering the industry's boom-and-bust cycle. He finds the company's valuation highly attractive and is increasingly interested, but does not state a qualifying forecast for its share price or investment return.
Narratives
Kevin is impressed by SK hynix's long-term customer agreements, custom high-bandwidth memory strategy, cash generation and exposure to AI-related storage demand. He regards its estimated valuation as deeply discounted for cyclicality and expresses investment interest, while stopping short of an explicit share-price or return forecast.
Key Arguments
- More than half of 2027 and 2028 production capacity is reportedly committed to long-term agreements, with discussions about extending contract periods.
- Custom memory products and deeper customer relationships could make demand more predictable.
- Advanced high-bandwidth memory consumes more wafers, so higher capital spending does not necessarily imply proportionately higher bit supply.
- Operating cash flow and free cash flow are rising despite increased capital expenditures, supporting shareholder distributions.
- AI demand is spreading across hyperscalers, sovereign projects and nontechnology enterprises, reducing dependence on individual customers.
- Kevin expects AI inference and agent workloads to increase NAND and SSD demand; SK hynix has exposure to both memory and storage.
- His tentative ADR calculation implies roughly six times forward earnings and a PEG ratio around 0.3, which he considers a substantial cyclicality discount.
- He thinks earnings growth could exceed the estimates he reviews, but this is an earnings outlook rather than a stock-return prediction.
Risks acknowledged
- Memory has historically experienced shortages followed by excessive capacity and collapsing prices.
- Contract expirations could leave investors exposed to renewed cyclicality.
- Kevin still believes DRAM supply could expand sharply.
- Management's explanation for restrained supply is a bullish company argument rather than independent confirmation.
- Analyst forecasts include an earnings contraction in 2029.
- The valuation depends on an ADR conversion and calculations Kevin explicitly says require fact-checking.
- He has not purchased any memory or storage investments.
NVDANvidiaKevin contrasts Nvidia's flexible GPUs with custom ASICs that perform specialized tasks but can lose usefulness when requirements change. He also discusses changes to memory specifications as an engineering and supplier-demand issue, without forecasting Nvidia's share price.
Key Arguments
- General-purpose GPU flexibility can justify a premium over specialized ASICs.
- Custom ASICs may lose value rapidly when workload requirements change.
- Memory capacity should be balanced against bandwidth constraints to avoid unused resources.
- Customized high-bandwidth memory could better match individual chip designs and customer requirements.
Risks acknowledged
- Customers may reduce memory specifications when additional capacity provides little benefit.
- Custom ASIC development presents an alternative to general-purpose GPUs for specialized workloads.
GOOGLAlphabetKevin uses Google's Workspace AI upgrades as a concrete example of AI monetization through recurring per-seat charges. He sees these upsells as supporting the broader case that AI adoption can produce revenue and cash flow, without making a directional forecast for Alphabet shares.
Key Arguments
- Workspace offers an additional AI access subscription that can be enabled for individual users.
- Higher usage limits and access to additional AI capabilities create incentives to upgrade.
- Per-seat pricing can generate substantial recurring revenue when deployed across large organizations.
- His observations support management's broader claim that AI infrastructure spending is translating into commercial adoption.
Risks acknowledged
- He says Google has been left behind in parts of the AI race.
- His Workspace observations illustrate a monetization mechanism rather than establishing a stock-price outcome.
Hedges & Caveats
- Kevin previously viewed memory stocks skeptically because of cyclicality and speculative momentum.
- Long-term agreements may defer rather than eliminate oversupply risk, particularly when contracts expire.
- He treats management's earnings-call commentary as a sales pitch that requires skepticism.
- He retains a thesis that DRAM supply could expand substantially.
- A slowdown in AI investment could undermine the boom; labor-market weakness remains a concern.
- He asks viewers to fact-check his ADR conversion and earnings-per-share calculations.
- He states that he owns no memory or storage investments.