AI | OpenAI is now reportedly generating around $20 billion in ARR and has secured roughly $1.4 trillion in long-term data center commitments. At the same time, the company is backing Sam Altman’s new venture, Merge Labs, which is working on brain-computer interfaces and memory prosthetics, so the long-term roadmap for AI now includes “talk to your spreadsheet” and “talk to your hippocampus.” Yep, that's a Neuralink competitor. Anthropic, meanwhile, is shipping faster than ever. The company rolled out interactive Claude apps integrated with tools such as Slack, Figma, Canva, Box, and more, allowing users to message colleagues, edit files, and trigger workflows directly within Claude, adding an interface layer on top of the model so it can interact with work rather than just chat logs. That said, only 12% of U.S. workers now use AI every day on the job, with roughly a quarter using it a few times a week and nearly half touching it at least a few times a year. A big jump from 2023 and a reminder that we’re still early in the S-curve despite how loud the AI noise feels. The Anthropic launch has put pressure on Altman and OpenAI, and it’s rumored that the company is considering a Q4 IPO to position itself ahead of Anthropic.

IPO WATCH | The early frontrunner for “hottest filing of the szn” might be…Barnes & Noble. After losing over $1B in market value from 2014 to 2019 and posting a $125.5 million loss on $3.7 billion in revenue in FY2018, the big-box bookseller looked cooked as Amazon hoovered up roughly half the market. Then Elliott Management bought it for $683M in 2019, swapped in new leadership, and basically ran a back-to-the-indie-playbook turnaround: fewer toys and random merch, more autonomy for local store managers, plus bolt-on deals for beloved independents like Denver’s Tattered Cover and Bay Area staple Books Inc. Now Elliott is reportedly prepping a combined Barnes & Noble/Waterstones IPO for as early as the back half of 2026, after opening dozens of new stores and driving roughly $400M in profit on around $3B in sales. On the opposite end of the vibes spectrum, Elon is cooking up a “megaverse” listing. SpaceX and xAI are in talks to merge ahead of a blockbuster IPO that could raise tens of billions and value the combined entity in the trillion-plus range, bundling rockets, Starlink satellites, the X social platform, and the Grok AI model into one everything-company. The idea is to lean into space-based compute and AI infrastructure as a single story for public investors, rather than trying to float xAI on its own alongside OpenAI and Anthropic. Another IPO candidate we have our eyes on is Canva, which is quietly one of the cleanest late-stage IPO setups in tech. The company is at about $3.3B in ARR, with a $42B valuation based on its latest employee stock sale, 240 million monthly active users, 27 million paid seats, eight consecutive years of profitability, and 800 million AI interactions per month. It has a new heavyweight CFO, has moved its corporate domicile, and is widely expected to file in 2026.

VENTURE CAPITAL | VC this month felt like a split-screen between industrial-scale capital and the craft side of the job. On one end, a16z just locked down another $15 billion across new vehicles, pushing firm AUM over $90 billion and representing nearly 20% of all U.S. VC raised last year. A16z is basically an asset class at this point, not a firm, with dedicated pools for everything from American Dynamism to infra to games. On the other end of the spectrum, a lot of the best writing in the industry right now is quietly hammering the same point: outperformance starts with origination, not with “winning” hot deals once they’re already in the bidding arena. The math is simple but under-discussed: your portfolio is constrained by the base rate baked into your funnel, so even a small improvement in the quality of the opportunities you see does more for long-run returns than a massive (and probably illusory) jump in your “picking skill.” That shows up in how the best emerging managers are behaving. Instead of camping out on Signal chains and Demo Days hoping to out-maneuver bigger brands, they’re building weird, opinionated systems to tilt the pool in their favor: communities of future founders rather than current ones, structured programs for young technologists, or research-heavy theses that start with scientific “sleeping beauties” and then work forward to company formation. Others are rethinking cold inbound, treating the inbox as a design problem rather than a chore, building lightweight filters, partner-led reviews, and brand surfaces that attract the sort of illegible, non-consensus founders who never make it into a mega-fund’s pipeline. All of this sits under a simple idea: it’s easier to sit in a different information flow than to be the smartest person in the same one.

BILLONS FOR BREX | Capital One acquired Brex this month for $5.15B in a cash-and-stock deal, the biggest bank–fintech acquisition to date and a meaningful liquidity event in an otherwise choppy exit market, even if it’s less than half of Brex’s $12.3B private peak. Brex started life as a very specific wedge (corporate cards for startups) and expanded into a broader spend stack. Despite the haircut to valuation, there was a clear split between those who made money and those who didn’t. YC’s original $120K check turned into roughly $100M. 800x return. 110% IRR across 9 years. Their follow-on through YC Continuity adds another $500M on ~$40M invested. Series A investors from 2017 achieved 80x returns at a 64% IRR. Series B in early 2018 made 12x at 39% IRR. These are the returns that make venture pencil as an asset class. Series C in late 2018 made 2.75x at 15% IRR. Then the 2020-2021 cohort. Series C+ investors got 1.3x at 4% IRR. Series D investors from Tiger and Greenoaks received their capital back with a 0% IRR. In a world where most late-stage 2021-vintage investments are underwater, breakeven is a win.

AI WISTLEBLOWING | A viral Reddit post blew up this month from an alleged food delivery app engineer claiming the company was basically rigged against customers and drivers. The “whistleblower” said the app slowed normal deliveries to make priority deliveries appear faster, adjusted fees in the background, and held certain orders so paid upgrades always seemed worthwhile. The big accusation was an internal “desperation score” for drivers, tracking how often and how quickly they accepted bad offers so the system could keep sending them worse ones. The post went viral on Reddit and Twitter. People started playing “guess the app,” and reporters rushed in to verify it was real, to the point where DoorDash CEO Tony Xu had to publicly say, “This isn’t us” and the company put out a blog post titled “How DoorDash is different.” Then it all fell apart: the whistleblower turned out to be an AI hoax. When a reporter pushed for proof, the whistleblower sent over an 18-page “internal” memo called “AllocNet-T: High-Dimensional Temporal Supply State Modelling,” supposedly from Uber’s “Marketplace Dynamics Group, Behavioural Economics Division,” complete with watermarks, charts, and dense math. It read like a real product spec, an AI system that scores drivers' desperation, discusses “greybulling” regulators, and even floats using Apple Watch data and ambient audio (crying, arguments, etc.) from drivers’ phones to decide when to underpay them. The reporter later learned the doc was AI-generated, the Uber badge image was AI-edited, and actual rideshare engineers said no one at a real company would ever compile every illegal thing they do into a clean, 18-page PDF. 😂 Uber's COO dropped a subtweet telling people not to believe everything they read online. It’s now easy for one bored person with an LLM to spin up something that looks like a leak, long technical docs, fake credentials, semi-plausible internal jargon and waste a ton of journalistic and public attention in the process. A decade ago, the effort required to create a fake like this would have been nearly impossible; now, the cost is near zero, and it takes less than an hour to impersonate an evil Uber Eats SWE.

$450M HOT DOG | One of the few companies actually delivering investor liquidity this month was Nathan’s Famous, which was acquired by Smithfield Foods in a $450 million all-cash deal announced last week. The brand’s story goes way back. Nathan’s started as a 5-cent hot dog stand on Coney Island in 1916, launched with a $300 loan. After expanding across New York, the founding family sold the business to investors in the 1980s, and it’s operated publicly ever since. Smithfield, which has held the rights to produce and sell Nathan’s-branded products in the U.S., Canada, and at Sam’s Clubs in Mexico since 2014, will now acquire all outstanding shares at $102 each. Like nearly every food company, Nathan’s has been feeling the inflation squeeze; its cost of branded products climbed 27% year over year, and a recent SEC filing noted a 20% jump in the average cost per pound of hot dogs. Still, the business has held up well, reporting more than $148M in sales for FY2025. Not bad for a brand that started with a 5-cent dog… maybe Joey Chestnut should’ve asked for some equity.🌭

EquipmentShare raised $747M in their IPO this month! Redbud GP’s Willy & Jabbok rang the opening bell as ES was minted as COMO’s latest publicly traded company. 🏗️

Chicago-based startups closed lots of Series A’s this month

Lots of rounds from Ohio-based startups were closed this month:

Cytotheryx, based in Rochester, Minnesota, raised a $60M Series A

Minneapolis-based OTTO Sport AI closed a $16.5M Seed round

Indy-based Opendate wrapped a $14M Series A led by High Alpha

When raised a $10M Series A led by 7wire & ManchesterStory

OpenAI Startup Fund credits

Anthropic’s startup program

AWS, GCP, and Azure startup tiers

Stripe Atlas + partner credits