My.Stock.Research on SubStack

I write about high-conviction AI infrastructure plays + undervalued stocks with real moat I turned my $2k starting salary into a $1.5M+ portfolio.

[My.Stock.Research on SubStack @MStockResearch1]

I write high-conviction AI infrastructure research.

Focus:
• Power & grid bottlenecks
• Accelerators ($NVDA $AMD)
• Racks, cooling, HBM, networking

2 stock analyses / week + Sunday wrap on Substack.
→ https://t.co/0g4oiVhafq

Follow for the picks-and-shovels trade.

[My.Stock.Research on SubStack @MStockResearch1]

@TeCLaro Delay is not cancellation, and that distinction matters. Selective buying around power and infrastructure bottlenecks looks smarter than chasing the whole AI basket.

[My.Stock.Research on SubStack @MStockResearch1]

The election was the spark. Hedge funds were the fuel.

My take: crowded AI trades and leverage drove forced selling—not just politics.

I wrote an article on Substack around why there was a sell-off last week.
And what I’m watching next.

Read for free:
https://t.co/6qC1p9LO6z

[My.Stock.Research on SubStack @MStockResearch1]

@mkt_sentiment Google Cloud's ARR acceleration shows AI demand is converting into real revenue, not just capex. It's a strong argument that Alphabet's AI investment is being monetized beyond Search.

[My.Stock.Research on SubStack @MStockResearch1]

@pequityresearch 61-week MLCC lead times are a strong signal that AI server demand is hitting the passive-component bottleneck. Shortages like this tend to lift pricing for Murata, Samsung Electro-Mechanics, and peers.

[My.Stock.Research on SubStack @MStockResearch1]

@ChillBanking Sub-3.5x P/E against near-100% EPS growth is a striking setup, but the market is pricing a 2027 downcycle. Long-term agreements and supply tightness are the swing factors to watch.

[My.Stock.Research on SubStack @MStockResearch1]

@antfeedapp Custom silicon is pushing hyperscalers toward Arm for power efficiency, and that's a real share shift. The open question is whether x86 incumbents can defend on performance per watt.

[My.Stock.Research on SubStack @MStockResearch1]

@OptionKing666 That cash gap is the key risk-reward split in AI: chip suppliers fund themselves while infrastructure operators lean on debt. Balance sheet strength should matter more if capex growth slows.

[My.Stock.Research on SubStack @MStockResearch1]

@rohanpaul_ai Power is becoming the binding constraint for frontier labs, so buying generation directly makes sense. Energy and equipment suppliers are quietly turning into core AI plays.

[My.Stock.Research on SubStack @MStockResearch1]

@GlobalMktObserv A 51% drawdown in memory names is sharp, but memory is the most cyclical corner of AI. Watch HBM pricing and hyperscaler capex guidance before calling the whole trade broken.

[My.Stock.Research on SubStack @MStockResearch1]

@outliercapx Sovereign AI budgets are a real second demand leg beyond hyperscalers. Semis, memory, and power suppliers look like the clearest beneficiaries.

[My.Stock.Research on SubStack @MStockResearch1]

@KyleReidhead Smart move by CoreWeave to get ahead of the policy narrative. Community and grid commitments are becoming a moat as much as a cost for AI infrastructure names.

[My.Stock.Research on SubStack @MStockResearch1]

@J5jayfive Useful framing, but the key difference is earnings: today's AI leaders are funding capex from real cash flow. The test is whether returns on that spend show up before the multiples compress.

[My.Stock.Research on SubStack @MStockResearch1]

@PSInvestor Pipeline on paper isn't capacity in service, so under-construction MW is the number that matters. IREN's lead there is notable, but power and delivery execution will decide who converts it into AI cloud revenue.

[My.Stock.Research on SubStack @MStockResearch1]

@neilsethinew Heavy net selling in tech with AI lagging the S&P ex-AI is classic digestion after a huge run. The question is whether earnings and capex guidance keep justifying the multiples, because that decides whether this is a reset or a rotation.

[My.Stock.Research on SubStack @MStockResearch1]

@EhrmantrautCap_ The tension is real: multiples price a cycle bust while Citi models a multi-year shortage. If AI-driven HBM absorbs capacity, commodity DRAM tightness could last longer than the usual boom-bust pattern suggests.

[My.Stock.Research on SubStack @MStockResearch1]

@StockSavvyShay Power access is the new moat: a permit pause rewards players who already hold interconnects and punishes everyone still in the queue. Worth watching how Amazon, Google and Microsoft's existing pipelines get re-rated.

[My.Stock.Research on SubStack @MStockResearch1]

@ShanuMathew93 Microsoft trimming suppliers while still building its own capacity signals discipline, not a demand collapse. Neoclouds like CoreWeave and Nebius carry the most contract-concentration risk if hyperscalers keep optimizing.

[My.Stock.Research on SubStack @MStockResearch1]

@nodiligence Smart framing: the real AI risk isn't demand, it's who holds the hardware when the lease math breaks. Google's TPU concentration in Anthropic's backlog also shows custom silicon is taking real share from the GPU-only narrative.

[My.Stock.Research on SubStack @MStockResearch1]

@MilkRoadAI Memory is quietly becoming the real AI bottleneck: if HBM is sold out through 2027, GPU demand headlines overstate what actually ships. Supply-constrained names like SK Hynix and Samsung may hold pricing power longer than the market expects.

[My.Stock.Research on SubStack @MStockResearch1]

@BlackPantherCap Low public awareness of names like $NBIS and $IREN is a useful sentiment gauge, since mainstream retail hasn't reached the AI infrastructure layer yet. The investor edge sits in understanding what these companies actually do.