Deep Dive
Record week closes strong on AI spending thesis
Markets ended July 6 on a high note, with the Dow adding 150 points and the NASDAQ up over 1%, capping a record-setting week. The S&P 500 momentum index sits at the 96th percentile against 40 years of history — the best second quarter in a midterm year since 1950. Technology led the charge with Broadcom up 4%, Tesla up nearly 6%, and Meta up 3%, while semiconductor stocks clawed back from early-week losses that saw names like Lam Research drop 20% in three days. The driver remains consistent: corporate spending on AI infrastructure is trickling through as revenue and profit growth for suppliers. Sanu Vargas, chief macro strategist at Carson Group, framed this as a wave unlikely to break before earnings season, describing pullbacks as consolidation rather than trend breakdown. Even if valuations stretch into bubble territory, he argued investors must ride momentum because the macro backdrop — strong capex investment, resilient consumer spending, and accommodative Fed policy — supports continued deployment through year-end 2026.
Consumer holding up despite wage pressure and inflation
Households are maintaining spending by tapping savings accumulated from a 156% surge in the S&P 500 and 50% jump in home prices over the past 6.5 years. Nominal consumption growth reached 7% annualized in April-May despite inflation headwinds, while nominal GDP sits at 6%. Real consumption remains below 2% and real wage growth is flat to slightly negative, but strong balance sheets are propping up the consumer. Vargas noted this dynamic isn't sustainable forever but will likely hold through 2026, buying time for inflation to pass through the system. The Fed, he said, will not raise rates through the end of the year — a stark contrast to markets pricing in a 25% probability of a July hike. The central bank is betting household balance sheets and corporate investment will cushion any inflation overshoot rather than forcing immediate tightening.
US-Iran deal becomes the market's fulcrum
Jake Connelly zeroed in on the geopolitical risk dominating the second half of 2026: the US-Iran peace agreement. He called it the single biggest market risk because every other macro concern — trade policy, AI capex appetite, Fed flexibility, midterm outcomes, and China's demand cycle — hinges on whether this deal holds. If the truce breaks down, oil prices spike, tariffs return, and the Fed faces pressure to tighten rather than accommodate, all while strangling the AI investment cycle. A sustained peace, conversely, normalizes oil prices, allows Persian Gulf fertilizer and metals exports to resume, and strengthens Asian economies dependent on semiconductor trade. The conflict has already depleted US arms stockpiles at enormous scale: over 1,000 Tomahawk missiles at $1.5 million each, plus hundreds of surface-to-air missiles, all requiring 4-5 year replacement cycles. Defense stocks are riding tailwinds from both inventory replenishment and the shift toward AI-powered autonomous systems as the true military advantage.
AI hardware race heats up in inference, not training
A Harvard dropout founder of an AI chip inference startup revealed a breakthrough in low voltage inference technology that achieves 80-90% FLOPS utilization versus the industry standard 30%, largely due to thermal throttling limits in current chips. The company has landed 1 billion dollars in contracts and is shipping its first product in summer 2026. By running inference at roughly half the voltage of competitors, the technology can serve large AI models at 200-500 tokens per second for roughly one-eighth the cost. He acknowledged Nvidia's dominance in training chips but insisted specialized inference hardware is essential as companies deploy at scale. The company opened its first factory in Taiwan and plans multiple additional sites globally. This segment underscores a broader theme: while Nvidia remains supreme for training, the inference opportunity — where models actually serve traffic and generate ROI — is fracturing into a competitive market requiring custom silicon.
Fed shifts toward opacity under new leadership
Kevin Worsh took over as Federal Reserve chairman with an explicit thesis: the Fed should talk less. Neil Irwin explained that constant communication — 70 people in the FOMC room doing about a dozen speeches per week combined — creates policy lock-in, preventing the Fed from nimbly responding to changing conditions. Worsh has appointed Mervyn King, former Bank of England governor, to co-chair a communication task force aimed at keeping Fed decisions off the front page and back in the business section where bond traders can parse them. This represents a sharp break from the post-2008 era of forward guidance and transparency. The theory: less pre-telegraphed decisions reduce the sense that policy is on rails, allowing the Fed room to surprise markets and adapt. Critics note this may create volatility around meetings, but Worsh seems willing to accept that trade-off for flexibility — a key advantage in an environment where inflation isn't cooperating with models.
Software AI monetization promises miss; real wealth elsewhere
John Deuchcci analyzed why Service Now and Salesforce were upgraded despite skepticism about their AI prospects. The upgrades came not because these companies will lead in AI but because current valuations assume perpetual 5% annual declines — an assumption too pessimistic. Service Now and Salesforce will likely persist and maintain relevance, but they won't drive outsized returns because adding AI to existing products doesn't justify significant price increases. Service Now initially hoped a 60% price uplift for ITSM with AI but is realizing a 30% uplift and settling for roughly the same price as before. Real wealth will come from entirely new applications that wouldn't exist without AI — products like Uber, which in 1999 nobody imagined would become a killer internet app. Current AI pricing based on tokens is unsustainable when costs exceed hiring humans, forcing models to normalize. This echoes a broader insight from Mark Pincus: founders often fail by holding onto losing products due to sunk cost fallacy. Success requires intellectual honesty about what's winning and ruthless focus on what customers actually value.