Ben Horowitz and David Solomon: The Sweetest Macro Spot in 40 Years

Summary

• Solomon sees one of the strongest macro setups in his “40-odd years” in markets for financial and investable assets. Significant and continuing fiscal expansion—including the Big Beautiful Bill that started in 2026—meets a rate-cutting cycle, deregulation and a capital-investment supercycle he describes as like something never seen before; the 4 largest companies contributed 1% to GDP growth with their $400 billion of spending last year. This “cocktail of stimulus” is hard to slow, although prices remain 25% to 30% higher and geopolitical risk has risen.
• Confidence has moved strategic transactions from an automatic “no” to “maybe,” setting up what Solomon predicts could be “the biggest M&A year in history.” He also expects a bigger IPO year as substantial private companies finally enter the pipeline. Horowitz agrees on activity but flags unresolved FTC uncertainty, especially toward tech, which might redirect deals into IP transactions.
• Goldman’s strategic problem is simultaneously scale and stable funding. Its $1.9 trillion balance sheet compares with JPMorgan’s $4.5 trillion; Solomon’s shorthand is that “when JPMorgan’s six, we’re gonna have to be at least three and a half.” Goldman now has about $500 billion of deposits—including over $200 billion through its digital platform—funding about 40% of the firm after having none 15 years ago.
• a16z’s rise rests on treating venture as a product for founders and then designing that product to scale. Horowitz recalls that “the best time to raise money is when nobody has money,” then describes a venture market that might expand from approximately 15 technology companies reaching $100 million in revenue annually toward 150. In 2025, about 18.3% of all U.S. venture capital raised was raised by a16z.
• AI is turning technological leads into capital-intensive races rather than durable software moats. The old “Mythical Man-Month” meant a company like Google could not simply use 1,000 engineers to wipe out a startup that had figured out a product with 7 or 8 people, but Horowitz argues that proprietary data plus enough GPUs can now solve “almost any problem. It is magic.” Companies reaching over $100 million—and in some cases $1 billion—in under a year may need public capital simply to keep competing.
• Enterprise AI’s largest payoff may be wholesale process redesign, not incremental assistance. Goldman spent $6 billion on technology last year but wanted to spend $8 billion; Solomon says finding $2 billion of efficiency would finance that additional investment without lowering returns. Its One GS 3.0 program begins with 6 processes, but success requires top-down pressure because employees are being asked to “take away their empire and do their empire differently.”
• Horowitz frames crypto and AI regulation as long-horizon U.S. competitiveness questions. After the GENIUS Act/stablecoin bill became law, a16z’s more important remaining crypto priority is the Clarity Act establishing market structure; on AI, its principle is “don’t regulate math,” regulate harmful applications. It also opposes 50 separate state regimes and restrictions on statistical training over copyrighted work when the training does not reproduce the work.

Deep dive

1. Goldman needs public-company scale without surrendering partnership

• Solomon describes Goldman’s old private partnership as “mutual agency”: partners built businesses, the collective reset ownership each cycle, and everyone went out again. Going public in 1999 supplied permanent capital for global expansion; without it, he says Goldman might resemble Lazard rather than today’s firm.

• The cultural bridge is a partnership class: every 2 years, roughly 450 people become Goldman partners, with compensation correlated to the whole enterprise. Yet Goldman also had to accept that “you can’t be a public company and not grow” or operate without top-down strategy making “1 plus 1 plus 1 plus 1” exceed the arithmetic.

• His CEO lesson after years advising others: “owning the responsibility” is fundamentally different from giving advice. The CEO must own growth and direction while scanning for strategic risks that could make the institution less relevant, less successful, less important or less competitive over the next 10, 15 and 20 years.

• Scale and funding dominate that scan. Goldman’s $1.9 trillion balance sheet trails JPMorgan’s $4.5 trillion; Solomon says Goldman will need at least a $3.5 trillion balance sheet when JPMorgan reaches $6 trillion. Goldman now has about $500 billion of deposits, including over $200 billion through its digital deposit platform, funding about 40% of the firm versus zero deposits 15 years ago. Ten years ago, Goldman was the world’s largest wholesale funder—a position Solomon says was “not one of” the things worth leading.

2. a16z turned founder service into scale—and scale into obligation

• Starting in 2009 drew “what are you, stupid?” criticism, but Horowitz’s retrospective is blunt: “The best time to raise money is when nobody has money.” Investors habitually chase high markets and retreat from low ones, giving contrarian entrants their opening.

• Because entrepreneurs ordinarily choose only “top tier” firms, a newcomer could not easily replicate the reputation Sequoia earned through Apple, Cisco, Yahoo and Google. a16z instead built a better founder product—brand, power, access and operating support designed to let founders remain CEOs rather than be replaced.

• Marc Andreessen’s 2011 “software is eating the world” thesis suggested that the roughly 15 technology companies reaching $100 million in revenue in a given year might become 150. That broke David Swensen’s basketball-team model of 5 or 6 investors, forcing a16z to scale coverage while keeping no more than 5 or 6 people involved in any deal.

• Having reached about 18.3% of U.S. venture capital raised in 2025, Horowitz invokes Andy Grove’s instruction: “If you’re the leader of an industry, then the growth of that industry depends on you.” He connects that responsibility to crypto policy, American dynamism and competition with China.

3. A four-part stimulus cocktail reopens the transaction machine

• Solomon’s “sweet spot” combines significant and continuing fiscal stimulus—including the Big Beautiful Bill that started in 2026—monetary easing with probably “a couple more” cuts, a capital-investment supercycle he describes as like something never seen before, and a deregulatory unwind. The 4 largest companies’ $400 billion of spending contributed 1% to GDP growth last year, making the economy unusually difficult to slow.

• Even if one talks about inflation falling from 9% to 3%, Solomon says Americans experience everything as 25% to 30% more expensive. He can name “a hundred things” that might derail markets, including a multipolar geopolitical order whose risk of producing a growth-slowing problem is “not high” but much higher than in recent decades.

• Two supports keep the market advancing: Solomon says the president “marks to market” daily and adjusts quickly when the market moves in the wrong direction, as around last April’s short-lived speed bump; meanwhile, the market is pulling forward AI productivity expected over the next 1, 2, 3 or 4 years.

• M&A confidence has shifted from “whatever the question was, the answer was no” to “the answer is maybe,” underpinning Solomon’s prediction that this could be the biggest M&A year in history. Horowitz’s pushback is FTC uncertainty: deals may become IP transactions instead. Horowitz also expects a lot of IPOs, including some out of necessity as companies grow rapidly, although going public means accepting that “of course you’re going to get sued.”

4. Crypto and AI policy have become national-competitiveness fights

• Horowitz presents crypto as more than financial infrastructure: it could define internet property rights, creator-led business architecture and stakeholder capitalism. He says the last administration effectively banned the sector through enforcement and “debanking,” with a16z receiving Wells notices despite being private.

• The GENIUS Act/stablecoin legislation is now law; a16z’s more important remaining priority is the Clarity Act establishing market structure. A token might represent a Pokémon card, stock certificate or dollar, requiring rules that distinguish them instead of treating everything—including an artist’s NFT—as a security.

• On AI, Horowitz insists, “The model is the model. It is a mathematical model”—not a sentient being. His proposed boundary is to regulate applications such as theft, bank intrusion or a robot that shoots somebody, while preserving the underlying mathematics: “Don’t regulate math.”

• The immediate concern is 50 separate state laws that startups cannot feasibly satisfy. Copyright is next: a16z wants models allowed to build statistical models over copyrighted works without reproducing them, arguing that weaker training access would disadvantage the U.S. because Horowitz says China does not respect copyrights.

5. AI’s enterprise payoff requires dismantling old processes

• Goldman’s first layer is getting tools, models and applications into employees’ hands so they can experiment and become more productive, powerful and consequential for clients. Uptake is accelerating, but unlike a company unconstrained by regulation, Goldman cannot say, “This is great. Let’s try it”; every tool must undergo a substantial regulatory-clearance process.

• Solomon is more interested in reimagining operating processes than adding incremental assistance. Efficiency is not solely about reducing people or costs—it creates capacity for growth investment while preserving annual accountability for returns.

• Goldman spent $6 billion on technology last year, though Solomon wanted $8 billion; $2 billion of process efficiency would close the gap without producing the hundreds of basis points in lower returns that the extra spending alone would cause. The first 6 One GS 3.0 redesigns could create “super significant” capacity, but must be driven top-down against internal empire protection.

• a16z is automating disliked work and putting all its hard firm and portfolio data into a Databricks data lake; Horowitz says customer support for it works “fantastic.” On agentic investing, Horowitz says models work from available facts, while the biggest portfolio-changing events can be completely new and unexpected. Solomon agrees that such events can be incorporated quickly once they happen, but wonders whether models based on the information held by underperforming investors will produce something different.