Michael Dell – Invest America Act Becomes Law, AI Talent Wars, Compute Demand, Market Update | BG2

2025-07-10 - 79 min - source - Read full transcript

Brad Gerstner (host)Bill Gurley (host)Michael Dell

Key insights

The Invest America Act, signed into law July 4, 2025, seeds a $1,000 S&P 500-invested account for every American child at birth, funded further by family, employers, and philanthropists.

Only children born after January 1, 2025 receive the government's $1,000 seed automatically; the ~65 million kids already under 18 are eligible to open accounts but must be signed up in a one-time enrollment campaign. Families can add up to $5,000/year tax-free and companies up to $2,500/year per recipient tax-free; funds compound in the S&P 500 and cannot be withdrawn until age 18.

invest-america-act

The program's fiscal cost is small: about $3.7B/year (3.7 million births x $1,000), roughly 1/100th of 1% of national revenue, and Gerstner argues it becomes revenue-positive in 20-30 years via capital gains taxes paid when accounts are eventually cashed out.

Gerstner frames the comparison as roughly equal to annual US foreign aid to Afghanistan and Nigeria combined, arguing it is a budget-priorities question rather than a deficit driver, while still describing himself as a longtime believer in a balanced-budget amendment.

invest-america-act

Because bipartisan negotiators could not agree on targeting contributions by household income, the bill instead allows geo-targeting by zip code (groups of 5,000+ people), letting donors and companies direct money toward lower-income communities indirectly.

Gerstner cites Milken Institute research finding that low-income households save at rates similar to higher-income households once they have an account, and that having an account correlates with higher high-school/college graduation, business formation, and homeownership, and lower incarceration.

invest-america-act

Meta's AI talent war (the $15B Scale AI acqui-hire, hiring Alexander Wang, Nat Friedman and Daniel Gross, and $75-100M pay packages) is possible because Zuckerberg is founder-controlled and can spend against a percentage of market cap the way Google and Apple, with the same cash reserves, structurally cannot.

Gerstner argues board and public-market pressure prevents non-founder-controlled companies from making bets this large and volatile; Zuckerberg is reportedly risking around 1% of Meta's market cap to reboot around AI after Llama 4 fell short competitively.

ai-talent-wars

Only a handful of companies, maybe 5-7, can sustain frontier/superintelligence-level AI talent competition, because it requires $10-40B in annual revenue to fund; Dell believes OpenAI and Anthropic clear that bar but the list of durable competitors is short.

OpenAI's roughly 2,800 employees are effectively all part of the frontier effort. Dell notes real startups cannot compete for this talent at all, and predicts large incumbents will generally get smaller in headcount while redeploying AI-driven profits into elite, separately-managed 'super intelligence'-style teams to avoid company-wide pay-fairness blowback.

ai-talent-wars

Dell's AI server business scaled from about $2B in shipments two years ago to roughly $10B shipped last year, with $12.1B in new AI orders in Q1 2025 alone and a backlog exceeding $14B; server/networking revenue grew 58% year-over-year.

Dell attributes the growth to heavy systems-engineering work beyond Nvidia's reference designs, supply-chain execution at 100,000+ GPU cluster scale, and Dell Financial Services helping fund deployments; Dell shipped the industry's first GB300 systems, delivering to CoreWeave and an undisclosed second customer.

ai-compute-buildout

Dell argues current AI infrastructure investment is still undersized relative to the productivity gains it could unlock: a 10% productivity improvement on the $114T global economy (two-thirds of which is services) implies roughly $10T of value, meaning annual AI investment should be on the order of $2-4T, well above actual spending.

Dell says roughly only 10% of large companies have figured out how to capture AI productivity gains today, with common results in the 10-40% range for those that have; the other 90% remain confused about how to apply it.

ai-compute-buildout

Dell is 98% confident AI's productivity impact will exceed both the PC and the internet because it touches essentially all knowledge work and expands the economic pie by creating new categories of work, not just cutting costs in existing ones.

He compares it to the difficulty of predicting new jobs 20 years ago at the dawn of the internet, and expects the gains to be expansionary for education, health, and societal outcomes broadly, not merely a cost-reduction exercise for existing firms.

ai-compute-buildout

Custom ASICs are gaining share specifically among the largest hyperscalers and model companies that can afford to optimize their own workloads, but this affects a small number of very large customers rather than the broader market, leaving Nvidia well positioned overall.

Dell notes on-prem and colocation deployment is also growing because many enterprise customers do not want to run their own data centers, and that most corporate AI use cases run fine on smaller, open-source models rather than the largest frontier models.

ai-compute-buildout

Gerstner flipped from being 'as negative as I've been in 10 years' early in 2025 (fearing $2T in Navarro-style tariffs) to bullish, citing 85% of S&P 500 companies beating earnings, tax predictability from the reconciliation bill, a Fed on hold, and tariff-related uncertainty largely digested by markets.

He notes the market shows heavy dispersion: Tesla down over 20% and Apple down 15% on the year even as the Nasdaq, S&P, TSMC, Nvidia, Microsoft, Oracle, Booking.com, and Uber sit at or near all-time highs, with the VIX back near 15-16 and the 10-year yield mid-range at 4.2%.

market-outlook

Both Dell and Gerstner warn that policy missteps, specifically the removal of the AI moratorium from the reconciliation bill (opening the door to roughly 70 state-level AI laws) and the slow pace of new AI-diffusion export licenses, risk squandering America's structural AI lead.

Dell argues for aggressively diffusing the full American AI stack globally, including allowing deprecated Nvidia chips to compete against Huawei inside China to preserve developer mindshare, using the MTOPS/PlayStation anecdote to illustrate how export controls on widely available consumer tech produce unintended consequences without achieving their goal.

market-outlook

Books referenced

Play Nice But Win - Michael Dell - Michael Dell's memoir, brought up by Gurley; Dell narrated the audiobook himself and argues authors should record it themselves to convey emotion and intonation

Media referenced

Jim O'Shaughnessy's podcast - podcast - Gurley references a prior appearance where he detailed how late-stage private funding cycles enabled today's AI pay packages
All-In Podcast - podcast - Referenced for Trump's comment about wanting to staple a visa to every diploma, tied to the skilled-immigration discussion

Companies

Dell Technologies - Michael Dell's company; AI server/networking business grew 58% YoY, took $12.1B in AI orders in Q1 2025 alone against a $14B+ backlog
Broadcom - Michael Dell described as a major owner following the VMware spinoff
VMware - Spun off from Dell; referenced as part of Dell's history
Meta - Acqui-hired Scale AI for $15B, brought on Alexander Wang, Nat Friedman and Daniel Gross, and is paying $75-100M annual packages to win the AI talent war
OpenAI - Cited as one of the few frontier labs (with Anthropic) capable of counter-spending against Meta's talent raid; ~2,800 employees, all effectively in the frontier race
Anthropic - Cited as one of the small number of companies with enough scale to compete for AI superintelligence talent
Google - Cited as having the same cash-generation advantage as Meta but unable to make founder-controlled, high-risk talent bets
Nvidia - Described as 'in a great spot'; reference designs require heavy additional engineering from partners like Dell to become deployable clusters
AMD - Named alongside Nvidia in the discussion of custom ASICs gaining share among large model companies and hyperscalers
Apple - Cited as having Meta's cash advantage but not making the same AI talent bets due to lack of founder control; also had an employee poached in the AI talent war
Microsoft - Cited as an example of a large company shrinking headcount while reaccelerating revenue via AI productivity gains
Uber - Cited as an early Invest America corporate matching partner and as an example of AI-driven margin expansion without headcount growth
Oracle - Cited as an Invest America corporate matching partner and among companies at all-time highs
Salesforce - Cited as an Invest America corporate matching partner
T-Mobile - Cited as an Invest America corporate matching partner
iHeartMedia - Cited as an Invest America corporate matching partner
DoorDash - CEO Tony Xu cited as publicly supporting Invest America
TSMC - Cited among companies trading at all-time highs despite market dispersion
CoreWeave - First customer to receive Dell's GB300 systems
Foxconn - Named as a competing ODM that Dell has out-executed on AI server delivery, per Jensen Huang's comments
Quanta - Named as a competing ODM that Dell has out-executed on AI server delivery
Hugging Face - Cited as the hub for the proliferation of open-source and smaller models used for on-prem enterprise AI
NextGen Personal Finance - Nonprofit (led by Tim Ranzetta) that Gurley funds; pushing financial literacy curricula, now required in 29 states, positioned as complementary to Invest America

Techniques and frameworks

Negative cash conversion cycle / build-to-order (FIFO Q) - Gurley's 1990s research thesis on Dell: building to individual customer order with ~6 days of inventory created a structural 200bps gross-margin and ROIC advantage over competitors carrying 90 days of inventory
Loan-to-value framing for national debt - Dell argues the US federal deficit/debt should be measured against the total value of US assets (~$200T+ vs ~$36T debt), not treated as a solvency crisis; the real issue is a spending problem, not a loan-to-value problem
Zip-code geo-targeting for philanthropy - Because household-income targeting couldn't get bipartisan agreement, Invest America uses zip codes (groups of 5,000+) as a proxy so donors can direct contributions toward lower-income communities
Golden age of margin expansion - Gerstner's framework for how AI lets companies reaccelerate topline growth while holding or cutting headcount, expanding margins broadly across the economy

Summary

Brad Gerstner and Bill Gurley hosted Michael Dell over the Fourth of July weekend for a wide-ranging conversation that opened with the newly signed Invest America Act. Gerstner, who spent years pushing the bill, and Dell, an early CEO Council supporter, walked through the mechanics: every American child under 18 (roughly 65 million kids) becomes eligible for a private, S&P 500-invested savings account, with children born after January 1, 2025 receiving an automatic $1,000 seed from Treasury. Families can add up to $5,000/year and employers up to $2,500/year per child tax-free, and a list of major companies (Dell, Uber, Nvidia, Oracle, Salesforce, T-Mobile, iHeartMedia) have already committed to matching. Because bipartisan negotiators couldn't agree on income-based targeting, the bill instead lets donors geo-target by zip code, which both hosts frame as a new philanthropic platform. Gerstner walks through the cost math (about $3.7B/year, a tiny fraction of federal revenue) and argues the program will eventually be revenue-positive through capital gains taxes, while Dell separately makes the case that the federal deficit is better understood as a "loan-to-value" problem against the total value of US assets rather than an imminent solvency crisis.

The conversation pivots to the AI talent war triggered by Meta's aggressive moves: the $15B Scale AI acqui-hire of Alexander Wang, the hiring of Nat Friedman and Daniel Gross, a wave of poaches from OpenAI, Google, and Apple, and reported $75-100M annual pay packages. Gurley traces the roots to the private-market funding cycle that let companies like OpenAI and Anthropic burn billions with investor tolerance unusual for public companies. Both hosts and Dell agree the deeper reason Meta can make this bet is founder control: Zuckerberg can risk roughly 1% of Meta's market cap on a reboot in a way that non-founder-controlled Google or Apple structurally cannot, given board and public-market constraints. Dell predicts the field of companies able to sustain frontier-level AI talent competition is small, maybe five to seven, requiring $10-40B in annual revenue, and expects most large incumbents to shrink overall headcount while carving out elite, separately-managed "super intelligence"-style units to avoid company-wide compensation blowback.

Turning to his own business, Dell details Dell Technologies' AI server ramp: from roughly $2B shipped two years ago to about $10B last year, $12.1B in new orders in Q1 2025 alone, and a backlog over $14B, with 58% year-over-year growth in the server and networking segment. He credits heavy systems engineering beyond Nvidia's reference designs, supply-chain execution at 100,000+ GPU cluster scale, and Dell Financial Services' deployment financing, noting Dell shipped the industry's first GB300 systems to CoreWeave. Dell argues current AI infrastructure spending remains undersized: applying a conservative 10% productivity gain to the $114T global economy (two-thirds services) implies roughly $10T of value, which should justify $2-4T/year in AI investment, well above what's actually being spent today. He states he is 98% confident AI's productivity impact will exceed both the PC and the internet because it touches essentially all knowledge work and expands the economic pie rather than merely cutting costs, though he estimates only about 10% of large companies have figured out how to capture the gains so far.

The episode closes on a market update and a policy warning. Gerstner describes flipping from his most bearish stance in a decade earlier in 2025 (fearing large-scale Navarro-style tariffs) to bullish, pointing to 85% of S&P 500 companies beating earnings, tax predictability from the reconciliation bill, and a Fed on hold, while noting sharp dispersion between mega-cap AI winners at all-time highs and laggards like Tesla and Apple down double digits on the year. Both Dell and Gerstner flag that policy could still undercut the moment: the removal of a federal AI moratorium opens the door to roughly 70 fragmented state AI laws, and AI-diffusion export licenses remain stalled despite the repeal of the Biden-era diffusion rule. Dell closes with the MTOPS/PlayStation anecdote, arguing export controls on widely available technology often backfire, and both men make the case for aggressively diffusing the American AI stack globally, including easing restrictions on deprecated chips competing against Huawei inside China, to preserve America's technology lead.

Notable Quotes

"Cash is king. Everything else is an opinion." - Michael Dell

"It's far bigger. It's far, far bigger. I feel 98% confident that it's far bigger than the PC, than the internet." - Michael Dell

"He's got the world's biggest printing press shooting out billion dollar bills. He's not relying on the beneficence of venture capitalists." - Brad Gerstner (on Zuckerberg)

"If they don't move quickly to reimagine their businesses given all this technology, they will be destroyed by new companies that come in with a totally clean slate." - Michael Dell

"You guys think you're going to control the access to this thing, or little things that move easily, you're kind of fooling yourselves." - Michael Dell (recounting the MTOPS/PlayStation Pentagon anecdote)