Here are my updated portfolios positions. Will be updating them regularly as best as I can. Most put sales are cash secured.

MS: HBM revenue is set to triple to ~US 200 B in 2027. $NVDA still the largest consumer of HBM in 2027 but it competes with $GOOGL, $AMZN, and $AMD as their share grows faster.

Largest companies by CY3Q26E operating profits. Same companies arranged by FY2027E P/E.

MS: Spending by largest cloud companies. Cloud capex reaches US 1.6 T by 2028e, but growth slows from 97% to 12%.

BTO $NBIS Jan’2028 170/200 risk reversal yesterday

BofA: We expect $NVDA to maintain ~70%+ AI accelerator share over time. $AMD, $AVGO, $MRVL

S&P 500 sector performances YTD

S&P 500 and Nasdaq 100: Percentage of stocks above their key moving averages

BofA: Our operator-by-operator analysis suggests global Al/cloud data center capacity could increase from ~41GW in CY25 to ~219GW by CY30E, with annual additions accelerating from ~22GW in CY26E to ~45GW by CY30E. Hyperscalers remain the largest contributors, while neocloud and sovereign deployments become increasingly material over time.

UBS: $MSFT Azure and M365 commercial cloud segments revenue estimates. 12-month PT 675.

JPM: AI capex increasingly relies on external funding and funding requirements are expanding - US 5.5 T of AI capex over 2026-2030. Can AI generate sufficient economic returns to sustain the investment cycle?

The age of easy money is fading, and investors aren’t ready for what’s next theglobeandmail.com/investin…

JPM: The AI compute mix continues to evolve toward heterogeneity, with custom XPUs gaining share of the accelerator TAM (XPUs rising from ~32% of units in CY25 to ~42% in CY26E and ~53% in CY27E) as hyperscalers scale TCO-advantaged internal programs. AVGO and MRVL remain the best-positioned custom silicon beneficiaries, with AVGO continuing to add XPU customers and lengthen multi-year XPU commitments across its hyperscaler base, while MRVL's custom ASIC and optical franchises continue to scale on strong design win momentum. On the merchant GPU side, NVDA continues to execute on the Vera Rubin ramp with POs secured from every major hyperscaler/AI cloud/system OEM, while AMD continues to execute on the MI4xx roadmap and Helios platform ramp.

JPM on AI/accelerated compute: Infrastructure demand continues to outstrip supply, with consensus cloud (hyperscaler + neocloud) capex for Y26/CY27/CY28 now at $950B/$1.42T/$1.55T (+97%/+47%/+9% Y/Y) and visibility continuing to push further into the decade on the strength of multi-year hyperscaler commitments and backlog build. The sustainability debate has shifted from whether AI demand is real to whether supply can keep pace. Even as hyperscalers continue to add capacity, both industry and company checks continue to indicate that near-term supply is insufficient to meet demand across GPUs, ASICs, HBM, and advanced packaging, with GPU rental pricing dynamics and lengthening order books at NVDA, AVGO, and MRVL all pointing to the same conclusion. We expect commentary from the leading compute providers to continue to point to upside to CY27 (and beyond) order visibility, with sovereign AI, AI-native cloud, and enterprise demand continuing to broaden the demand base.