Weekly Research Briefing: Summer Drag

Heading into the summer, consumer spending momentum was very strong, with total credit and debit card spending rising 6.3% year-over-year (YoY) in June - the strongest growth in over four years - according to Bank of America internal card data. With gasoline prices falling, the increase in spending growth is almost entirely a discretionary story.

“I think we’re probably in sort of third inning if I -- that might be conservative, but we are starting to see changes in behavior. So the most notable now that we haven’t sized yet, but we’re watching closely is summer shutdown patterns are changing significantly. So whereas we would have had preplanned shutdowns, particularly in automotive, those are being canceled. Those are not -- or they’re not being announced as they would have been. So it’s still spotty, but it’s real. And so I think that’s probably the best indicator that we have.” - MSC Industrial Direct CEO Martina McIsaac

“AI is like the next-generation force multiplier for small businesses,” said van Zyl, head of Askari Defense Inc. in Atlanta.

Nationwide, economists say we’re seeing a second wave in the business formation boom that started in the Covid-19 lockdowns, when worried and laid-off workers rushed to start their own businesses. Only this wave appears closely correlated with AI’s growth, economists say, with the strongest gains in sectors amenable to AI usage, such as professional services.

Across all industries, the US Census Bureau projects 29,700 new businesses will form monthly within a year, based on initial IRS paperwork filed by entrepreneurs. That’s 17% more than formations projected a year ago, based on June data published Thursday.

The AI-friendly professional services sector (architects, lawyers and advertisers) is expected to see even faster growth, with more than 5,000 companies formed monthly over the coming year, a record in data going back to 2004 and 24% more than a year ago.

Federal Reserve Governor Christopher Waller said policymakers may need to raise rates in the near term if underlying inflation continues to signal broad price pressures.

“If we get another hot reading on core inflation this week, then the FOMC will need to consider tightening monetary policy in the near term,” Waller said Monday, referring to the central bank’s rate-setting committee, in remarks prepared for an event in New York…

“No matter how you cut it, or what measure you want to use, inflation is up this year,” Waller said. “At this point, I am concerned about the elevated pace of core inflation.”

He pointed to the version of the Fed’s preferred inflation gauge that strips out volatile food and energy components. The core personal consumption expenditures index hit 3.4% in the year through May. It began rising in January, Waller said, before the US-Iran war began, and “has steadily moved up.”

Stock-to-stock correlation inside the Russell 1000 just touched levels seen only a handful of times in ~40 years. A quiet index $SPX. A loud tape w/ winners and losers canceling each other out. Think alpha vs beta.

SALEM TOWNSHIP, Pa.—When land developers drove to Marilee and David Kiliti’s farm two years ago, the couple invited them into their home, sat them down and sliced up some deer bologna.

The men told the Kilitis that their 89-acre farm in this rural town of 4,000 might be worth more than $20 million. To Marilee and David Kiliti, that ridiculously high price sounded like a bunch of deer bologna.

The couple thought the fields where the family raised and butchered hogs would be lucky to fetch even a fraction of that. Unlike other parts of the state, there was no oil or natural gas in the ground.

But the developers had no interest in farming or fuel. The Kilitis’ land was worth a fortune—more than $22 million—as the future site of a data center…

The Kilitis are one of 96 families in Salem Township, an unassuming community in the northeastern part of Pennsylvania, who collectively sold about 1,700 acres to a single data-center developer, QTS, an arm of the investment firm Blackstone. The families sold for an average of $330,000 an acre and earned $5.5 million on average. The total sale price was $586 million.

San Francisco’s pandemic-era doom loop has been replaced by a land grab, driven in part by anxiety about how artificial-intelligence wealth will impact home buying.

“It’s definitely the Wild, Wild West,” said David Cohen, founder of City Real Estate. “The market has had a huge influx of new demand.”

Buyers are rushing to front-run anticipated IPOs from giants like OpenAI and Anthropic, local real-estate agents said. “For many people not in AI, the feeling is, ‘If we don’t get in before the big AI IPOs, we’re going to miss our window,’” said Kate Tomassi, an agent at Sotheby’s International Realty – San Francisco Brokerage.

In reality, AI money is already flowing, agents said. Wealth is rippling across the entire AI ecosystem, from hardware manufacturers to AI developers. Some employees at leading, still-private AI companies are already armed with liquid cash from secondary-market transactions and novel arrangements with local lenders.

“I’ve got junior guys who are selling secondary shares and buying $10 million homes,” says Phil Chen, a real-estate agent at Christie’s International Real Estate Sereno. Tomassi adds, “The money is real, and it’s a lot of cash.”

With the wealth created through the two massive AI public offerings coming down the pipeline, current and former employees of OpenAI and Anthropic could buy nearly one-third (29%) of all homes in San Francisco, where both companies are headquartered. That’s not just homes listed for sale–it’s 29% of all homes in the entire metro area…

SpaceX employees could hypothetically pool their IPO windfall to buy roughly 40% of all the homes in San Antonio, one of the closest major metros to the company’s Starbase, TX headquarters, according to a separate Redfin report. The total value of all homes in San Antonio was roughly $297 billion in 2024; we estimate SpaceX employees hold roughly $200 billion in equity at the midpoint of our 10–15% ownership estimate, which comes to $120 billion post-taxes.

Rather than betting on the next breakthrough technology in artificial intelligence, Novogratz wants to own the very infrastructure the industry relies on, by staking his claim in Texas.

That conviction about grid-connected power — as well as the turf and data centers built on top of it — is transforming Galaxy into an AI infrastructure landlord alongside its existing crypto businesses.

“It’s the simplest part: get power, find a good tenant, lease it, borrow the money and build it and then you’re a landlord,” Novogratz said. “For me, I want to get as much power as I can and I want to get it leased as quickly as I can because after that I am just a rent collector.”

Galaxy is not the only crypto company to pivot to AI during a prolonged rout in Bitcoin — far from it. A growing number of firms have been seeking fresh opportunities in AI as their native businesses have become less lucrative.

The crypto market has now erased more than half of its value in the past eight months, and bitcoin ETFs have seen $8 billion in outflows in the past eight weeks. Meanwhile, AI has become the new frontier. It’s the technology people are most excited about, and it has captured the same sense of disruption and possibility that crypto once represented.

The company said in a blog post on Monday that the site in Richland Parish, Louisiana — home to what will be Meta’s largest data center — will be a 5GW facility and cost over $50 billion. That’s higher than the $27 billion figure that was revealed in October, when Meta and Blue Owl Capital formed a joint venture to help with the buildout and management of the facility, originally planned as a 2GW data center.

Updated bottom up cost per GW and rising forward capacity expectations cause us to raise hyperscaler capex to $1.4t in '28 painting a path to 120 GW (from 30 in '25)… We have written extensively about the rising value, strategic importance and monetization optionality associated with access to compute. In effect, limitations on chips/racks, powered shells and other bottlenecks are stretching timelines from breaking ground to opening data centers to as much as 3 years. Furthermore, we also believe growing social backlash to data center development and political uncertainty into the '28 presidential election are causing hyperscalers to start building sooner to emphasize job creation. This adds even further to inflationary pressures.

The entire +33% jump in US IG supply YTD has been driven by three sectors: AI, M&A and banks. Away from those three categories, IG supply is mostly unchanged from 2025 (Exhibit 7, Exhibit 8). That makes sense, as downside risks to supply created by higher borrowing costs and strong earnings offset the upside risks to supply created by +$39bn increase in maturing debt in 2026 relative to 2025. That suggests these three sectors should continue driving incremental issuance volumes in 2H-2026 as well.

“We still need visibility for when AI revenues will start to fund much of this cash need. Expect more advertising plays from LLMs, Token prices have to decline to drive Enterprise adoption. Expect LLMs to chase more vertical profit pools, legal, life science, expecting physical AI companies. Pure models will continue to see arbitrage with open source touching 30% usage, depth will create a better moat, breadth will commoditize. It’s not a demand problem - ‘it’s a monetization problem’. Silicon valley has always built product with intensity and the market has funded adoption years. This time it might just be too big and the market may not have capacity to fund everyone. ‘Darwinian moment for AI providers?’” - Palo Alto Networks CEO Nikesh Arora

AI-generated content appeared across all social media platforms in our data set. The average AI rate across all scanned items was 13.8%, but specific rates varied by platform and item length. On four out of five platforms, longer content was more likely to be AI-generated than shortform content. Across all platforms, one in four longform items (25.72% of items over 250 words) were fully AI-generated.

Substack was an exception; there, the rate of fully AI-generated content remained fairly flat, and longer, more substantial posts were actually slightly less likely to be AI-generated compared to shorter ones.

LinkedIn was the most AI-saturated platform, where more than 40% of longform posts flagged as fully AI-generated. However, if we included mixed AI and human content, X/Twitter was the worst off: almost half of X articles were either fully AI-generated (23.9%) or AI-assisted/mixed (22.9%), with only 53.2% of X articles flagging as fully human-authored.

Americans are leaving some of the nation’s most flood-prone communities at an accelerating pace, suggesting that climate concerns are beginning to influence migration decisions alongside more traditional factors such as housing costs and quality of life.

High-flood-risk US counties — those classified in the top 10% for their share of homes that are very vulnerable to flooding — lost a net 63,357 domestic residents from mid-2024 to mid-2025, nearly double the outflow recorded over the previous 12-month period, according to an analysis released Wednesday by real estate brokerage Redfin. In 2024, flood-prone counties collectively posted a net population loss for the first time in five years.

Meanwhile, counties with relatively low flood risk gained nearly 70,000 residents on net, the largest increase since 2018…

“Affordability is still the primary reason people move, but affordability and climate risk are becoming more tightly linked,” Redfin Chief Economist Daryl Fairweather said. Rising insurance premiums and homeowners association fees, and maintenance costs associated with extreme weather, are increasing the cost of owning a home in flood-prone areas, she said.