TMTB Morning Wrap
Good morning. QQQs -1.5% as higher yields weight on sentiment this morning - 10 Year +6.5bps to 4.55% as the Moody’s downgrade — while not surprising — “serves to remind markets, which had become quite complacent and expensive in the last few weeks, that there is a serious fiscal problem that needs to be reckoned with” (VK).
Big week last week as QQQs sit 3% off ATHs as risk on ruled the day - just take a look at CRWV’s 50%+ move in 5 days.
This week, we have a few events: Computex is already underway as Jensen gave the keynote late last night (eastern time) - recap below. QCOM, Mediatek, and Foxconn CEO’s are today. MSFT’s Build Conference also starts today - expectations are muted with most focus on Copilot agents and potential AI features for Windows 11 and Copilot+PCs. Google’s I/O starts tomorrow and stock has rallied 10% into the event following AAPL’s Cue’s commentary a couple weeks ago, helped by some better 3p data on Friday from Yipit which called out an accel in the last two weeks and improving sentiment around AI releases. GOOGL is expected to tend Gemini across products and unveil “AI Mode” inside Search; a bigger model or fresh usage metrics could counter cannibalization fears, while even a whiff of an Apple tie-up would be incremental upside for the narrative. On Wednesday, GOOGL Marketing Live kicks off where advertisers should get new generative-AI creative and Measurement upgrades plus a clearer read on how AI Overviews and Performance Max are monetizing; any confirmation that ad yield matches classic Search would reassure bulls on the durability of Google’s core cash engine.
Let’s get to it…
NVDA: Jensen Keynote from Computex here
Key quotes:
“These so-called AI data centers are improperly described — they are in fact AI factories. You apply energy to them and they produce something incredibly valuable.”
“The opportunity has gone from a $1 trillion data-center market to an AI-factory industry that will be measured in trillions of dollars.”
“Grace Blackwell systems are in full production … coming online all over the place every single day. They’re already in CoreWeave, used by many CSPs, and you’re starting to see them ‘coming up from everywhere’.
“The more you buy, the more you make. That’s what factories do.”
“That single one-gigawatt AI factory will cost 60-80 billion dollars, and 40–50 billion of that is the electronics — the compute systems we make.”
“Humanoid robotics is likely to be the next multi-trillion-dollar industry, and the consumption of compute and data-centers will be enormous.
Summary:
Jensen highlighted that every region and industry will deploy AI-specific data centers, each costing $60-80 billion with roughly two-thirds going to compute hardware—implying multi-year, visibility-rich demand for GPUs, networking and HBM. The production rollout of Grace Blackwell systems is already under way at CoreWeave and major CSPs; an upgraded GB300 (1.5× inference, 2× network bandwidth) ships in Q3 ’25. The newly announced NVLink Fusion lets hyperscalers graft their own ASICs or CPUs onto NVIDIA’s switch fabric—broadening TAM while making rivals partial customers.
Beyond hyperscale training, Huang staked out fresh growth vectors likely to drive incremental revenue and attach‐rate software:
Enterprise AI: a rack-scale RTX Pro Enterprise/Omniverse server (8 Blackwell GPUs + 800 Gbps networking) aims to put “agentic AI” into corporate IT stacks; DGX Spark and a desktop DGX Station shrink a petaFLOP into wall-plug form factors, expanding the developer base.
Industrial & robotics: Omniverse digital-twin adoption by TSMC, Foxconn, Delta and others positions NVIDIA for recurring software and GPU sales as $5 trillion of new fabs and factories are designed in simulation; humanoid robots, trained with Jetson-Thor and “Groot 1.5” models, are pitched as the next multi-trillion-dollar compute sink.
Gaming: the GeForce RTX 50 launch was called the fastest in company history.
RDDT: Reddit downgraded to Equal Weight from Overweight at Wells Fargo
Wells Fargo analyst Ken Gawrelski downgraded Reddit to Equal Weight from Overweight with a reduced price target of $115 from $168, citing "more permanent" user issues as Google implements AI features in search. The threats are not only to Reddit, but all the companies dependent upon search traffic, the analyst tells investors in a research note. Ken expects "logged-out user declines" to pressure the multiple —he also cuts #s, forecasting ad revenues 6%/14% and EBITDA 2%/15% below consensus for '26/'27. Ken views logged-out users (55% of total) as "critical to extending ad reach" despite contributing only 15% of ad revenue, and warns that Reddit's data licensing business may be "incompatible with maximizing community value," suggesting the company should prioritize internal monetization over the "$150M high margin revenue stream" to support long-term growth
Third Party Data Roundup:
Keep reading with a 7-day free trial
Subscribe to TMT Breakout to keep reading this post and get 7 days of free access to the full post archives.