This was one of the most information-dense weeks of the year. BofA added Micron to its US 1 List. Intel reported its best quarter in 15 years. Alphabet confirmed Google Cloud grew 82%. Tesla beat on revenue. The Magnificent Seven lost $767 billion in two days. MU fell to $900 despite all of it. Intel reversed 6% on Friday after jumping 10% in afterhours trading Thursday. Kimi K3 dropped. Brent hit $100. The 10-year yield touched 4.71%. On Friday, the S&P 500 is down about 0.7% for the week and the Philadelphia Semiconductor Index is on track for a weekly decline of roughly 10%, its largest weekly loss in over a year.
The market went through nearly every emotion available, and arrived back roughly where it started, with one critical difference: investors are now repricing the timeline between AI capital expenditure and free cash flow in ways that will define sentiment well into Q3.
Monday-Tuesday (July 20-21): The Institutional Conviction Setup
The week opened constructively. BofA's addition of Micron to its US 1 List alongside a $1,550 price target and Morgan Stanley's simultaneous upgrade projecting DRAM prices rising 25% next quarter drove MU up 15.10% to $976.94 on Tuesday. UBS simultaneously published its view that the momentum unwind was nearing its bottom, noting that hedge fund positioning in semiconductor and software stocks had returned to April levels after one of the largest gross exposure reductions on record at roughly 5% of market value.
The analytical case was compelling. Micron trading at approximately 6 times fiscal 2027 consensus EPS of $154, against the S&P 500's 20 times forward multiple, represented a fundamental-to-price dislocation that three of the most followed institutional research teams were publicly flagging at the same time. Tuesday's move reflected that capital returning ahead of the week's earnings catalysts.
Wednesday (July 22): Alphabet Sets the Template and the Market Rejects It
Alphabet reported Wednesday after 1:00 PM PT with numbers that were operationally extraordinary. Revenue of $119.8 billion beat the $116.9 billion consensus. Google Cloud grew 82% to $24.8 billion, the fastest growth rate in the segment's history, expanding operating margin to 35.6% from 20.7% a year ago. Gemini reached 950 million monthly active users. The cloud backlog grew to $514 billion.
The stock fell 6.9% Thursday because the earnings call disclosed that full-year 2026 capex guidance had been raised again to $195 to $205 billion, free cash flow turned negative for the first time in company history at minus $5.9 billion, and CEO Pichai confirmed Gemini 4 is being trained as a much larger model requiring more scale. The market's verdict was specific: 82% cloud growth is not in question, but the willingness to fund $500 million per day of AI infrastructure without a visible FCF recovery timeline is no longer unconditionally supported at current valuations.
Tesla compounded the sentiment damage. Q2 revenue of $28.24 billion beat the $26.32 billion estimate, FSD subscriptions reached 1.48 million ahead of estimates, and the CFO signaled Q3 deliveries will be flat rather than down 8% as consensus models. But adjusted EPS of $0.33 missed the $0.50 consensus by 34%, Musk described 2026 as a "massive capex year," and the CFO confirmed capex will grow for another two to three years. Tesla fell 14.3% on the day, the worst single-session decline since April.
Thursday (July 23): Intel and the Counternarrative Nobody Noticed
While the Magnificent Seven was losing $767 billion, Intel reported Thursday after 1:00 PM PT. Revenue of $16.1 billion beat the $14.45 billion consensus by 12%. Non-GAAP EPS of $0.42 beat the $0.22 estimate by 91%. Data Center and AI revenue grew 59% to $6.3 billion. Foundry grew 31% to $5.8 billion. Intel 18A-P entered risk production. Q3 guidance midpoint of $16.3 billion came in above estimates. The stock jumped 10% in afterhours trading.
The GAAP result showed an $11 billion loss, but the mechanism was specific and non-operational. Intel holds shares in escrow for the Department of Commerce under CHIPS Act terms. When Intel's stock price rises, accounting rules require booking the mark-to-market increase in that liability as a loss. Strip out the $12.5 billion non-cash charge and Intel earned $2.2 billion. The company lost $11 billion on paper because its own stock became more valuable.
Intel's result did something neither Alphabet nor Tesla managed: it demonstrated that AI infrastructure revenue can be generated at a capital intensity level the market can still value. Intel's 59% Data Center and AI growth came on $20 billion of annual capex, not $205 billion. That capital efficiency contrast was the week's most underreported analytical data point.
Friday (July 24): Kimi K3, Geopolitics, and a Second Reversal
Friday opened with two simultaneous catalysts pulling in opposite directions. Chinese AI startup Moonshot unveiled Kimi K3, described as the world's largest open AI model with 2.8 trillion parameters, claiming performance rivaling Anthropic's frontier Fable model. The announcement revived the same open-source AI demand narrative that BofA had addressed in its Monday note, where Arya argued open-weight models with 1.4 terabytes of HBM per inference instance multiply rather than reduce memory demand. The market processed the Kimi K3 announcement as a headwind regardless.
Intel reversed sharply, falling 6% Friday after Thursday's afterhours jump, as the broader semiconductor sector sold off alongside it. MU declined roughly 8% back toward $900. The VanEck Semiconductor ETF pulled back 3%. The move carried the technical signature noted in options flow analysis from earlier in the week: lower highs forming after a lower low, with the weekly fair value gap sitting meaningfully below current prices.
The geopolitical counterweight was a Reuters report citing three Pakistani sources that China and Pakistan are considering a path toward new US-Iran peace negotiations. Oil pulled back on the news, and the Dow briefly turned positive with Apple contributing a 3% gain on reports of strong Siri AI feature adoption in the iPhone upgrade cycle. The S&P 500 and Nasdaq remained in negative territory heading into the final minutes.
On Friday, the S&P 500 is down roughly 0.7% for the week, the Nasdaq is tracking a loss of roughly 2.9%, and the SOX is on pace for a weekly decline of approximately 10%, which would be its worst weekly performance in over a year.
What This Week Actually Established
Five things are structurally clearer after this week than they were before it.
The AI demand is not in question. Google Cloud at 82% growth, Alphabet's $514 billion cloud backlog, Intel's 59% Data Center and AI revenue growth, AMD's 12 gigawatt customer commitments from OpenAI and Meta, and Tesla's strongest order backlog since 2023 all point in the same direction. The AI infrastructure buildout is real, growing, and accelerating by every available measure.
The market is repricing the capex-to-cash-flow timeline. Alphabet's negative FCF quarter, Tesla's multi-year capex commitment, and the pattern of guidance raises without accompanying FCF recovery visibility have shifted institutional sentiment from "trust the investment" to "show us the return." That repricing has further to run ahead of Microsoft, Meta, and Amazon reporting next week.
Capital efficiency is being rewarded. Intel's 59% AI revenue growth on $20 billion capex outperformed every Magnificent Seven name on capital efficiency this week, and it is the only company in the AI hardware complex that is trading above where its earnings print implied it should trade.
The momentum factor unwind is the largest in 27 years by Goldman Sachs' measure, and it is interacting with genuine macro headwinds including Brent at or near $100, 10-year yields approaching 4.71%, and a Fed meeting on July 28 to 29 with a non-trivial hike probability.
The technical picture for the semiconductor complex is fragile. MU's pattern of lower highs after a lower low, the SOX entering and then briefly exiting bear market territory, and the absence of a volume-exhaustion capitulation signal all suggest the consolidation phase is not over. SK Hynix reports July 29. Microsoft and Meta report next week. Those events are the next fundamental anchors.
Explore more exclusive insights at nextfin.ai.
