Alphabet’s $80 Billion Raise Shows Public Markets Regaining AI Power

John Coogan framed Alphabet’s reported $80 billion equity raise as a sign that the AI buildout has changed the function of public markets. Citing James Mackintosh’s Wall Street Journal piece, “Alphabet’s Mega Fundraising Shows the Value of Being a Public Company,” Coogan emphasized the argument that “in AI, money talks”: the ability to tap stock-market capital matters again after a long stretch in which private markets could often finance large companies deep into maturity.

The claim was not that Alphabet needed cash like a startup. It was that the AI race has become capital-intensive enough that even the largest companies are raising and deploying money at industrial scale. Coogan pointed to researcher compensation, data-center construction, and the financing of losses while AI businesses mature as evidence that competition increasingly turns on access to money as much as on models or products.

He cited the claim that 61% of all venture capital last year went to AI. Jordi Hays said that “feels low,” partly because so many companies now describe themselves in AI terms and because “there was a lot of hard tech.” Coogan’s broader point was that AI has become difficult to discuss as a single sector: it touches sycophancy and consumer behavior, data centers, water and power, enterprise software, SaaS, and infrastructure.

The reported structure of Alphabet’s raise mattered. Coogan described it as a staged plan: Berkshire Hathaway buying $10 billion of shares at roughly a 6% discount to Monday’s close; another $30 billion through underwritten public offerings; and $40 billion in staggered common-stock offerings beginning in the third quarter of 2026. Because Alphabet was described as a roughly $4 trillion to $4.5 trillion company, the dilution was presented as modest: $80 billion is huge in absolute terms, but less than 2% of the company’s market value.

The unresolved question was why Alphabet would issue equity rather than rely only on debt. Coogan used Ben Thompson’s framing: debt is normally the preferred instrument when a company has confidence in the return on investment, because existing shareholders keep more of the upside, though the company carries repayment risk. Equity removes debt risk but gives up a future share of profits.

That left two interpretations. One was that the equity raise could be a precursor to more debt, meaning investors still underestimate the scale of compute demand. The other, more cautious reading, was that Alphabet may be uncertain about the return on all the capital expenditure and wants shareholders to share risk as well as upside. Coogan said that if there is not a substantial debt issuance later, the bearish interpretation may look more plausible.

Compute costs were the practical hinge. Even at the level of inference cost, Coogan said, labs are seeing “dollars per task,” with large sums flowing into data centers. But when those tasks are compared with other ways of getting the same work done, he said the results can still show positive return on investment. In that view, the AI buildout is justified by productivity uplift, provided the work being automated or assisted is valuable enough.

The Wall Street Journal’s larger capital-markets point was that Big Tech’s borrowing is already large enough to affect bond markets. Coogan cited the Journal’s claim that bond investors believe hundreds of billions of dollars of Big Tech debt are pushing up yields for other borrowers and even affecting government bond yields. Alphabet, Microsoft, Amazon, and Meta were described as major bond issuers as they increase spending.

For the riskiest and least settled AI investments, however, equity is a more forgiving instrument. Coogan highlighted the Journal’s argument that shareholders do not have to be repaid, so if AI takes longer to make money or never makes money, a company financed by stock can wait longer than one constrained by debt covenants and repayment schedules. Investors may be unhappy, but the company has more room.

That is why Alphabet’s public-company status becomes a competitive advantage. Coogan said Alphabet is one of the few companies able to raise this much cash without “tanking its stock,” because of its search business and its credibility with Wall Street in new ventures. He argued that Google’s side projects have historically been mocked or discounted by the market, but that some are beginning to show power-law potential. Waymo was his main example: one successful outcome there could outweigh many discontinued apps, experiments, and “April Fool’s jokes.”

Coogan also noted a more adversarial interpretation circulating around the raise: Alphabet may be absorbing investor liquidity and AI demand before future IPOs from companies such as Anthropic, OpenAI, or SpaceX. Hays offered the simpler alternative: “you should probably raise capital when it’s cheap.” Coogan agreed that capital has been pulled from other market sectors and has to go somewhere; the “latest and greatest technology” is the obvious destination.

The Journal’s historical comparison was to 19th-century railroads: stock markets exist, in part, to funnel money from millions of savers into giant projects. For the last 25 years, private capital reduced the urgency of that function. AI’s cash needs, in this telling, may be beyond the private markets’ capacity.

Coogan preserved the bear case. Equity raising at record prices can also look like opportunistic selling near a top. But his main read was that AI has made capital access strategic again, and Alphabet is showing how valuable it is to be able to raise public-market money quickly, at scale, with limited dilution.

Berkshire Hathaway’s reported $6.8 billion agreement to acquire Taylor Morrison Home Corporation put the scale of AI spending in relief. John Coogan noted that roughly comparable headline deal sizes produce very different outcomes: one buys an entire home builder, while the other buys only a tiny slice of Google. He called the housing deal interesting and important, but said it looked like “total peanuts compared to the AI buildout.”

Tyler, a producer on the show, added that both deals were small compared with Berkshire’s cash position, which he said was most recently $397 billion. Coogan used that to keep the Berkshire narrative in perspective: even with new CEO Greg Abel making large moves, Berkshire remained “cash Chad.”

The Taylor Morrison acquisition was presented as Abel’s first major deal as Berkshire’s new chief executive. Coogan quoted Warren Buffett’s brief reaction to the Journal: “He has launched.”

Jordi Hays read the underlying thesis as a long-term belief in American housing. The deal was described as increasing Berkshire’s exposure to a housing market in its fourth year of weak sales. High mortgage rates, job-market uncertainty, and the rising cost of living have kept many buyers on the sidelines. Builders have had to offer incentives, including paying part of buyers’ mortgage costs, to move inventory. Builder confidence was described as low; single-family home starts fell 9% in April, and a third of builders said they cut prices in the prior month.

The optimistic case rested on supply shortage and deferred demand. Hays cited analysts who say the United States has a housing shortage of more than 4 million homes, meaning new homes need to be built. They expect buyers to return when mortgage rates come down from recent highs and pent-up demand is released. Berkshire agreed to pay a 24% premium to Taylor Morrison’s closing stock price of $58.58. Coogan reacted by calling the price an “incredible bargain.” He then cited Tony Avila, chief executive of Builder Advisor Group, for the view that the deal was attractive because the actual value of the builder’s home portfolio belied its lagging stock price.

Taylor Morrison was described as a safer bet than a generic homebuilder because it focuses more on the higher end of the market, which has held up better. A meaningful part of its business is built around buyers upgrading to nicer homes rather than entry-level buyers under the most pressure. The company also participates in build-to-rent communities: single-family homes constructed specifically for rental use.

That build-to-rent exposure carried a policy risk. Coogan cited a congressional proposal that would have forced build-to-rent developers to sell properties within seven years of building them, though House lawmakers removed that provision in an attempt to preserve the sector. The deal also fit a larger consolidation trend in residential construction, alongside the reported AvalonBay Communities and Equity Residential merger in multifamily housing.

The housing discussion returned to consumer psychology. Coogan said he was encouraged by the claim that more than 75% of young renters still think they will someday own a home. He said online sentiment might suggest a lower number, but that Gen Z homeownership looks different once one separates expensive coastal cities from non-coastal markets. Some people move to San Francisco or similar cities to be “in the mix,” rent longer, and delay family-planning decisions; affordability problems remain, but the desire to own has not disappeared.

Anthropic’s confidential draft S-1 filing became a way to explain how confidential IPO filings became standard practice. Liz Hoffman wrote that the confidential IPO process began as a 2012 rule change intended to ease small companies with less than $1 billion in revenue into public markets, and asked, in reference to Anthropic, “what are we even doing here?”

A confidential IPO filing does not mean a company goes public in secret. It means the company submits a draft S-1 registration statement to the SEC for private staff review before the prospectus appears publicly on EDGAR. Once public, hedge funds, journalists, competitors, and anyone else can download it. The confidential period lets the company work through SEC comments while sensitive financial details remain private.

Jordi Hays called it a “smoke grenade.” John Coogan accepted the analogy, but argued that the practice has practical value. It allows regulatory problems to be addressed before the public filing, reducing the chance that a company has to pull back after the market has already seen its numbers.

Before the 2012 change, Coogan said, the S-1 became public early in the process. That was good for journalists covering IPOs, but it raised the stakes for companies. If something was wrong, or if the process stalled, the result could be interpreted as a failed IPO, damaging morale and signaling weakness. “No one wants to be running a company that publicly failed to IPO,” he said.

Coogan described the 2012 JOBS Act as creating “emerging growth companies,” or EGCs, with relaxed filing requirements. He said the threshold began at companies with less than $1 billion in revenue and was later adjusted to $1.235 billion. In 2017, he said, SEC staff under the first Trump administration expanded the confidential filing path to all issuers, not just EGCs. That meant larger companies could use the same process regardless of revenue.

He emphasized that the JOBS Act was Republican-driven but broadly supported, and that the 2017 expansion did not face strong opposition. The reason, in his telling, was that both sides of the market benefit from more IPOs. Private-market investors want liquidity. Venture capitalists want to return capital to limited partners and take victory laps when portfolio companies go public. Public-market investors want access to more names.

The 2017 context mattered because large private companies had become far bigger and more complex before listing. Coogan named Uber, Airbnb, DoorDash, and Palantir as examples of decacorns that were already past the billion-dollar revenue threshold but still difficult for public markets to value. Airbnb, for instance, was not simply comparable to a hotel chain because it was asset-light. Uber also lacked clean public comparables. Confidential filings gave the market and companies more room to prepare for novel business models.

Coogan said the process expanded again in 2025, under SEC staff in the second Trump administration, to include other financial offerings: new stock issuances, other classes of securities, spin-offs, follow-on financings, and other capital-markets transactions. His description was that companies could “test the waters” before launching a broader market process.

His conclusion was pragmatic. Companies are staying private longer because private markets are deep: large venture funds, crossover hedge-fund investors, hyperscalers, and strategic investors can provide enormous checks. That has pulled companies away from public markets. Exchanges and public-market investors do not like that. Confidential filings are one way to make going public less painful.

Coogan argued that the process does not eliminate investor protections because the traditional disclosures still have to be released before any public money changes hands. What it changes is timing, coordination, and the risk of public embarrassment. After looking into it, he said he did not “really have a problem with it,” while leaving open the question of whether listeners thought the practice should be rolled back.

Hays added two related notes, both kept as attributed claims. First, he said Joe Weisenthal was reporting that more than a thousand current and former SpaceX employees had banded together, ahead of what Hays called “the eve of the IPO,” to negotiate with wealth management firms for better pricing and access to tax-saving financial products. Second, Hays noted that Alibaba, Uber, and Meta all went public on Fridays. Coogan expected that if SpaceX did go public, the coverage would be a spectacle, especially if the imagery he attributed to its S-1 was any indication, with Starbase “in full force.”

OpenAI’s expansion was physical and product-level at once. John Coogan said OpenAI announced that it was breaking ground on Stargate Michigan, a 1-gigawatt data center using closed-loop cooling. OpenAI Newsroom’s post said the project would use water “at the rate of a typical office building,” create thousands of union jobs, and provide more than $40 million in free Codex credits for college, community college, and trade school students statewide.

The water claim mattered because data centers have become politically sensitive around water and power use. Coogan described OpenAI as “getting out ahead of the water FUD.” Jordi Hays turned the office-building comparison into a joke about Coogan’s bottled-water consumption, but the substantive point was that OpenAI was not just announcing compute capacity. It was pairing the data center with labor, education credits, and a resource-use reassurance.

The product update was Codex Sites. OpenAI said Codex can turn work ideas and plans into an interactive website or app that a team can explore, use, and share through a URL, rolling out first to Business and Enterprise plans before broader expansion.

For Coogan, this addressed a specific product benchmark: could he open an AI app on his phone, generate a website, and share a link with a friend? He compared that to the current portability of AI-generated text and images. A text thread can be shared. An image can be saved to the camera roll and sent around. But interactive software often remains trapped inside the originating app.

Meta AI supplied his example. One suggested prompt could “vibe code a video game,” and Coogan said it did a strong job creating a small game. But the share link remained inside Meta AI rather than becoming a normal hosted website. The missing layer was an actual hosting service. Codex Sites, in his view, made AI-created interactive artifacts more portable and therefore more viral.

Hays predicted “a bull market for simulators.” Coogan agreed, saying many people want to build from a phone, not only from a Mac mini or a MacBook Pro with the lid open. If users can create something interactive and send it around as a link, the output becomes part of normal communication rather than a demo locked in a tool.

The sharpest security example came from 404 Media’s Jason Koebler. Koebler wrote that hackers had been stealing high-profile Instagram accounts by “simply asking Meta’s AI support chatbot to change the email associated with the account they want to steal.” He called it a “shockingly easy” and “terrible flaw” tied to offloading support to AI.

John Coogan summarized the 404 Media report as a serious example of AI support risk: hackers asked Meta AI to give them access to high-profile Instagram accounts, and, according to Koebler’s reporting, it worked. Coogan said he assumed Meta was rolling it back, but treated the reported exploit as a warning about technical support systems that lack proper validation and access controls.

Koebler explained the reported method directly: attackers started a chat with Meta’s AI chatbot, said they wanted access to a specific account, and asked for a reset code to be sent to the attacker’s email address. According to Koebler, “Meta’s AI chatbot said, sure thing.” He said high-profile accounts targeted in the prior 24 hours included Barack Obama’s White House account, a Space Force account, and Sephora’s account. 404 Media had spoken with multiple users whose accounts were targeted and had joined Telegram groups where the method was being circulated.

Jordi Hays said he had seen people he knew with one-word usernames get hacked. Coogan’s concern was not that AI support is inherently impossible, but that the chatbot appeared, in Koebler’s account, able to trigger account recovery changes without segmentation or proper verification. In that failure mode, the support system becomes an automated privilege-escalation path: the attacker does not need to defeat a human process if the AI assistant can be persuaded to initiate the recovery flow.

The smaller AI items sharpened the contrast between useful assistance and poorly bounded automation. OpenAI’s Terence Tao advertisement offered the deliberate version of human-AI collaboration. OpenAI said AI can give researchers the freedom to pursue “crazier” ideas and that, for Tao, AI creates room to experiment, test unexpected paths, and discover what might otherwise remain out of reach.

John Coogan called Tao a “GOAT mathematician” and said he has spoken often about using models to flesh out work or build charts he might not otherwise build. The framing was not replacement. It was the “centaur” model: a human expert using models to extend exploration, test paths, and create intermediate artifacts that would otherwise take too much time.

That was the clean version of the AI product story. Codex Sites similarly mattered because it addressed portability: an AI-made app becomes more useful if it has a URL and can circulate outside the tool that generated it. The Meta account-recovery issue was the opposite failure mode. As Koebler reported it, an AI support workflow appears to have been given authority over account recovery without enough verification. The difference is not whether AI is involved; it is what authority the system has and whether the surrounding process can absorb mistakes or manipulation.

The more mundane consumer example came from Mark Gurman’s report that Apple was preparing an iOS 27 service for splitting bills at dinners and events by photographing a receipt and assigning items to friends. The feature would be part of Apple Wallet and Cash and would compete with Venmo and Splitwise. Jordi Hays called it “annoying super intelligence” that he would not use. Coogan preferred credit-card roulette, then turned even bill splitting into a game-theory problem: if the table is splitting evenly, each diner is incentivized to order the expensive steak, more drinks, and dessert to avoid subsidizing everyone else.

The point was narrow but important: AI features that work technically still have to survive the social or operational context where they land. A research assistant that helps Terence Tao explore more paths is one kind of product. A support bot with account-recovery authority is another. A receipt splitter may solve a payment problem while making dinner feel more transactional. The useful boundary is not the presence of AI, but whether the deployment matches the trust and behavior required around it.

A Joe Weisenthal post said “the price of whey is going bananas,” linking to Bloomberg. The excerpt described HelloAmino, a Canada-based baking and beverage company that uses whey protein in all 30 of its high-protein baking mixes. In early May, a supplier told the company it had run out. Founder Aelie Swift found another supplier, but it required importing whey protein isolate from the United States at a price 50% higher and expected to rise again. The new whey also changed the product: a different processing method dried out the baked goods. “Our pancakes came out like sawdust,” Swift said in the excerpt, and the company planned to reformulate with a different combination of proteins because whey had become too expensive to use as before.

John Coogan used the item to note how far the protein trend has spread into consumer packaged goods: cereals, pastas, baking mixes, salads, lunches, dinners. He said he had not personally found protein “sneaking into” many of the things he eats — “there isn’t protein in my Diet Coke” — but he recognized that protein-enhanced foods had become a category. Jordi Hays argued that the idea someone eating two or three solid meals with protein also needs protein snacks in between is “completely unnecessary.”

The item was lighter than the capital-markets discussions, but it echoed the same operating reality at a different scale. Demand narratives become supply-chain problems. A consumer trend toward high-protein foods does not stop at branding; it reaches ingredient availability, supplier quality, input prices, and product reformulation. HelloAmino’s problem was not merely paying more for whey. It was that a substitute ingredient changed the pancakes.