Editor’s Note
This week was not a referendum on whether AI is real. It was a referendum on who can pay for it, who can monetize it, and who is being priced as if the answer is already guaranteed. Across Big Tech capex, semiconductors, robotaxis, crypto, gold, Netflix, and leveraged markets, investors kept returning to one lesson: a great thesis can still be a bad trade at the wrong price.
AI Infrastructure Is Still the Center of Gravity
The strongest theme of the week was that AI spending is moving down the stack. Dan Ives framed the cycle as early, saying, “We’re still in the third inning of the AI revolution,” while also warning that China’s AI progress can create another “white knuckle moment.” His key distinction was simple: “It’s less about the models, it’s about the data.”
That distinction mattered as investors reacted to Moonshot’s Kimi models and another “Deep Seek” style scare. The bearish read is that cheaper Chinese models could pressure OpenAI, Anthropic, and U.S. hyperscaler economics. The more nuanced read is that cheaper models may expand usage, not reduce compute demand. Kimi’s owners reportedly said their servers were “cooked” because demand overwhelmed capacity. If lower AI prices create more enterprise use, the model layer may see margin compression while Nvidia, memory, semiconductor equipment, cloud infrastructure, and data-center suppliers continue to benefit.
Alphabet made the upside and cost visible at the same time. Q2 revenue was $119.8 billion, above the $117 billion estimate, with operating income up 30% YoY to $40.8B. Search revenue rose 17% YoY to $63.3B, undercutting the view that AI is already destroying Google’s core business. Cloud revenue came in around $24.77 billion to $24.8B, above the $22.46 billion estimate, while one source said Cloud surged 82% YoY and operating income tripled to $8.8B, with margins rising from roughly 21% to roughly 36%.
But the bill was enormous. Alphabet’s Q2 CapEx was $44.92 billion, slightly above the $44.15 billion estimate, and another source framed it as $44.9B against $39.1 billion of operating cash flow. Free cash flow was negative by about $5.8B or -$5.86B, described as Google “burning through cash” for the first time since going public decades ago. FY26 CapEx guidance reportedly rose to $195B - $205B from $180B - $190B, or in another summary from $190B to $205B. One post called it, in Portuguese, “A maior bolha de todos os tempos.”
The market wants proof that capex becomes durable returns. Google reportedly started the week at 356 and fell 11% despite 24% revenue growth, $120 billion in revenue, and about 16 times forward earnings. Wall Street followed the tape: Piper Sandler cut its $GOOGL target to $395 from $445, Wells Fargo to $411 from $418, Cantor Fitzgerald to $420 from $435, DA Davidson to $350 from $375, Raymond James to $400 from $425, and UBS to $379 from $400. Barclays raised its target to $425 from $405.
Semis Still Have Demand, But Not All Profit Pools Are Equal
Semiconductors remained the purest expression of AI conviction. Tom Lee argued that weakness in DRAM, memory stocks, U.S. semiconductors, Korean chip-related stocks, SK Hynix, and Samsung was painful but not thesis-breaking. Some names were around 20% cheaper, which Lee called “That’s the entry point.” He also said, “We would be buying any of those dips.”
SK Hynix’s 13% Nasdaq debut surge supported the supply-constrained view. Ives said memory demand-supply equilibrium may still be 2 years away and described the market as a “12 to 15 to one type of demand supply environment.” AMD also added fuel: Lisa Su said the AI accelerator market could reach $1.4T by 2030 inside a broader $2T compute market, while AMD unveiled Helios, a rack-scale AI system built around the MI450 accelerator. AMD may invest up to $5B in Anthropic, which may buy up to 2GW of AMD Instinct MI450 chips starting in the first half of 2027; another source said the first one gigawatt deployment is expected to begin in 2027.
Intel showed the turnaround-versus-price tension. Q2 revenue was $16.13 billion versus estimates around $14.43 billion to $14.48 billion, adjusted EPS was $0.42, and Q3 revenue guidance was $15.8 billion to $16.8 billion. Revenue rose 25% year over year from $12.86 billion, and Intel generated $7 billion in cash from operations. CEO Lipton said, “AI is driving unprecedented demand for compute,” and data center sales reportedly soared 59% last quarter. Yet Intel was still up about 171% year to date, and one valuation model, with the stock around $112, produced values of $14, $45, and $99.
JoAnne Feeney added discipline to the chip debate. He remained constructive on NVIDIA and Broadcom, citing security, stability, reliability, data handling, and custom infrastructure. But he was more cautious on Micron, even at around six times future earnings or 6.5 times forward twelve-month earnings, because memory remains cyclical. Asked whether memory is no longer cyclical, his answer was blunt: “I just disagree.”
Leverage, Valuation, and Reality Checks
The week’s market stress was not limited to AI. South Korea showed how a correct thesis can become a forced-selling event. Samsung’s Q1 2026 operating profit rose 756% year over year to 57.2 trillion won, while SK Hynix revenue rose 198% and operating profit rose 405%. At one point this year, Samsung was up over 500%, and SK Hynix was up over 1,000%. But Samsung and SK Hynix grew to nearly 60% of the KOSPI at the June high, while retail investors added $80 billion over six months and foreign capital pulled about $95 billion out.
Then the unwind hit. The KOSPI fell around 4.6% in one session, then more than 10% on “Black Tuesday.” Over 3 trillion won in investments were liquidated, 1.2 million accounts hit margin call thresholds, and 320,000 accounts were wiped out. Normally, about 2% of Korean margin accounts get force-liquidated; during the crash, that rose above 10%.
The U.S. has its own warning signs. The S&P 500 PE ratio is 32, the Shiller PE is 42 versus a long-term average of 17, and the Buffett Indicator is approximately 210% versus a historical normal around 100%. Margin debt was cited as up 54% year-over-year in one discussion and 55% year-over-year in another, with U.S. margin debt reaching roughly 4.5% of GDP as of June 2026, the highest level ever recorded.
Tesla, SpaceX, and robotaxis brought valuation back to earth. Waymo was running half a million paid rides per week, with more than 3,000 autonomous vehicles in 11 cities, while Tesla appeared to have roughly 200 cars across four cities. Texas DMV data showed 175 Tesla robotaxis versus 642 Waymo robotaxis in Texas. Between July 2025 and March 2026, Tesla robotaxis had 37 crashes causing property damage or injury, about one crash per 46,000 miles, worse than Tesla’s cited U.S. average of one per 174,000 miles for minor collisions.
Tesla shares fell 19% from 381 to 313 after Q2 earnings. Revenue grew 25% to $28.24 billion, but Q2 adjusted EPS was $0.33 versus the $0.51 estimate, gross margin was 16.8% versus 19.4%, and free cash flow was -$1.09B. Musk’s long-term conviction remains massive: “I think Optimus will be the biggest product ever.” But this market is asking for near-term proof.
Other Assets Are Being Repriced Too
Netflix became a cleaner valuation debate. The stock is down 44% over the last year, but the business still has 325 million global paid subscribers, 13% year-over-year growth, next-quarter guidance of 11.7% growth, and trades around a 22 P/E ratio and below 20 forward. The range of outcomes remains wide: with 10% average growth and a future P/E ratio of 20, intrinsic value was about 56; with 12-15% growth and a future P/E ratio of 25, value rose to about 85; with slower growth and a 15 P/E ratio, value could fall toward 30.
Gold also tested conviction. After reaching approximately $5,500 an ounce in early 2026, gold is closer to $4,000 an ounce, down about 26%. Yet over 25 years, the S&P 500 is up approximately 461%, while gold is up 938%. With U.S. debt at $39.6 trillion, central banks buying roughly 1,000 metric tons of gold per year for the past 4 years, and China buying gold for 20 straight months, the structural case remains alive.
Crypto’s issue was liquidity. Tom Lee cited monetary tightening, regulatory uncertainty, AI FOMO, and weak financial stocks. The bond market moved from pricing two Fed rate cuts to 1.6 hikes, effectively removing four cuts from 2026. But the June core CPI report, released on July 14th, showed a negative monthly core reading, giving crypto bulls a possible macro relief path.
Looking Ahead
Next period, the key watch is whether Big Tech can defend AI capex with cash-flow evidence, not just growth language. Alphabet, Meta, Microsoft, Amazon, Nvidia, AMD, Intel, and memory names remain central because the market is separating infrastructure winners from companies merely spending to stay relevant.
Also watch rates, oil, and leverage. Crude moved back above $90/barrel, the 10-year Treasury yield crossed 4.7%, Brent traded around $96, and tariffs of 10% to 12.5% on 60 countries added to the inflation debate. The week’s takeaway is simple and durable: AI demand remains powerful, but price, funding, and execution now matter again.