a16zSteven Sinofsky & Balaji Srinivasan on the Future of M&A, AI & Tech
CHAPTERS
- 0:00 – 4:13
Why M&A feels broken: IPO drought, deal blocks, and the Figma backlash
Balaji and Steven argue that U.S. capital markets have become hostile to tech: fewer IPOs since Sarbanes–Oxley and a tightened M&A window due to aggressive antitrust enforcement. They use Figma’s path (and the Adobe breakup fee) to illustrate how regulators claim credit for wins while ignoring downstream damage to startups and exits.
- •Post–Sarbanes–Oxley decline in IPOs and longer private-company timelines
- •FTC/DOJ scrutiny narrows M&A exits; examples of deals blocked with negative consequences
- •Figma’s IPO framed as succeeding despite (not because of) regulators
- •Critique of Lina Khan’s “victory lap” narrative and DC’s zero-sum incentives
- 4:13 – 6:04
Computing’s unusual origin story: a world-changing industry built without licensing
Steven reflects on how computing and software largely arose without the licensing and approvals common in other professions. The absence of early regulation created a fast-moving ecosystem that later collided with regulators who felt excluded from (and threatened by) tech’s expansion.
- •Software engineering and software sales historically required no licensing/approval
- •IBM antitrust as a rare early exception; otherwise minimal oversight
- •FCC/radio certification as one of the few practical barriers to shipping computers
- •Regulators’ discomfort with an economy-swallowing sector that grew “without hearings”
- 6:04 – 7:30
Network vs. state: why tech platforms became a political rival to government
Balaji lays out his “network versus state” framework: the internet became a primary actor that is upstream of elections, media, and commerce. As platforms gained de facto regulatory power (ratings, moderation, access), the state perceived a loss of authority and “struck back.”
- •Internet as an invisible but foundational layer for modern society and politics
- •Platforms increasingly regulate behavior (e.g., Uber as ‘taxi regulator,’ YouTube as ‘speech regulator’)
- •Conflict emerges when networks reach ‘state-level’ scale
- •Different incentives: consent-driven building vs. coercive allocation of power/status
- 7:30 – 21:13
Why antitrust struggles in software: market definition, HHI math, and disruption
Steven explains how antitrust tools were designed for tangible, distribution-constrained industries, not fluid software markets. They discuss how “market definition” can be arbitrary in tech, where products overlap and disruptive entrants don’t look like competitors until it’s too late.
- •Sherman/Clayton-era thinking assumes clear markets and measurable shares
- •HHI and market-share analyses depend on debatable market boundaries
- •Examples: word processors, email, phones by manufacturer vs. OS
- •Disruptive innovation (Christensen) arrives as ‘non-peer’ then becomes substitute
- 21:13 – 25:05
The regulator mindset and the ‘platform admin’ analogy—plus a numeracy gap
Balaji argues that many policymakers lack intuition for scale (money, orders of magnitude), causing poor decisions and narrative-driven enforcement. He adds that running massive platforms forces centralized defaults (flip-a-switch governance), which helps him understand more authoritarian instincts—though markets constrain platforms more than they constrain government.
- •Claims many state actors are selected for verbal over quantitative reasoning
- •Anecdotes about misunderstanding millions vs. billions and wealth concepts
- •Operating huge platforms requires defaults and centralized changes; not everything can be ‘consent/auctioned’
- •Key difference: platforms face competition/exit; governments often do not
- 25:05 – 31:44
M&A as a power-law bet: why deals are retconned after they succeed
They pivot directly to M&A economics: corporate acquisitions are often value-destructive on average, yet a few outliers transform companies. Regulators and commentators, they argue, ignore the risk at the time and later treat successes (YouTube, Instagram) as ‘obvious’ monopolist plays.
- •Corporate M&A studies often show net value destruction; success is rare
- •Power-law dynamics: a small number of deals drive most upside
- •Contemporary skepticism around YouTube and Instagram purchases
- •Retcon problem: nobody wanted the risk, everyone wants credit for the reward
- 31:44 – 41:27
When M&A works (and fails): size ratios, integration realities, and ‘make vs. buy’
Balaji and Steven discuss practical heuristics for successful acquisitions, emphasizing how difficult integration is and why mergers often fail. They break down internal big-company dynamics—hysterical memos, discovery risk, and the frequent mistake of buying #2/#3 instead of the category leader.
- •Heuristic: acquirer often needs to be ~100× larger; 10× deals can be ‘must-win’ but rare
- •Startup-startup mergers usually fail; acquisitions can work if incentives align
- •Big-company ‘make vs. buy’ pressures and internal politics
- •Common failure mode: buying cheaper #2/#3 and expecting ‘magic distribution’ to fix it
- 41:27 – 44:05
Transformative acquisitions beyond the headline: Microsoft FrontPage as a ‘DNA’ deal
Steven offers a detailed example of an acquisition that mattered more for capability transfer than direct product revenue. Buying FrontPage helped Microsoft internalize web editing/publishing, catalyzing broader shifts in how the company approached internet-era tooling.
- •FrontPage as a web ‘word processor’ plus publish-to-server workflow
- •Internal bidding war dynamics (Office vs. IE) as a barrier even before external competition
- •Value of acquiring talent/insight that changes company direction
- •Acquisitions as ecosystem-shaping, not just product-line additions
- 44:05 – 49:02
The new wave of ‘innovative’ AI deal structures: acquihire vs. the ‘acquifire’
Balaji explains why recent AI deals (Scale, Character, Inflection, Adept, Covariant, Windsurf) often avoid full acquisitions: antitrust risk pushes buyers to purchase teams and rights while leaving the company intact. He introduces ‘acquifire’—investors/employees get money via a structured payout, but the remaining entity loses the status of being acquired.
- •Antitrust fear incentivizes non-traditional structures instead of straightforward acquisitions
- •Acquihire: status/job continuity, limited company-level payout
- •Acquifire: large payout left behind (aligned to liquidation waterfall), but less public ‘acquired by X’ status
- •Need for explicit contractual planning (e.g., a ‘non-key man’ designated successor)
- 49:02 – 53:51
Windsurf case study: why status mattered as much as cash—and how optics spiraled
They walk through the Windsurf saga: Google wanted engineers, not a newly built sales org, leaving many employees behind with a cash-rich shell meant to wind down. Because the structure couldn’t be communicated clearly and didn’t confer acquisition status, a second deal with Cognition restored status but raised integration challenges (small acquirer absorbing a bigger org).
- •Google’s incentives: acquire top technical talent; avoid duplicative sales functions
- •Communication constraints from ‘not technically an acquisition’ created confusion
- •Employees left behind: money without the resume/status halo of ‘acquired by Google’
- •Cognition follow-on acquisition: status restored, but risky integration given size mismatch
- 53:51 – 56:23
AI platform shift and why tooling is getting ‘irrational’ investment (driving big deals)
Steven argues AI represents a platform shift akin to early PC consolidation, where tooling becomes strategically vital even if it’s not a huge standalone business. That dynamic helps explain seemingly ‘unreasonable’ prices for AI coding and developer-tool startups and the scramble by platform contenders to lock in the ecosystem.
- •Shift in tech nexus: from mobile/cloud toward AI as the organizing platform
- •Tooling is essential for platform adoption and gets funded beyond near-term ROI
- •Historical analogy: DOS-era tooling proliferation and ecosystem consolidation
- •Deal activity reflects platform competition more than simple financial modeling
- 56:23 – 1:02:40
Regulation as a cat-and-mouse game: ‘hacks’ to rules, from banking to antitrust
Steven places today’s deal innovations in a long history of businesses working around rigid rules, which then triggers regulatory backlash and new constraints. Balaji adds that sustained regulatory attacks can tempt firms to build lobbying power and go on offense, changing the character of the industry.
- •Sherman/Clayton origins and how business practices evolved around them
- •Examples of rule-workarounds: NOW accounts, telecom pricing innovations
- •Today’s ‘deal structure innovation’ as a predictable response to oversight pressure
- •Risk of arbitrary enforcement and escalating cycles between regulators and firms
- 1:02:40 – 1:13:52
AI backlash vectors: copyright, energy constraints, China’s open models, and US policy risk
Balaji outlines a multi-factor risk scenario where U.S. AI progress is slowed by lawsuits, power limitations, and restrictions on using Chinese open-weight models—potentially pushing innovation offshore as happened with crypto. Steven largely agrees on the strategic risk, framing China’s open releases as commoditizing U.S. strengths, while urging a nuanced view of copyright’s role in building the software industry.
- •Potential ‘perfect storm’: copyright suits + energy limits + strong Chinese open-weight models + US restrictions
- •Political backlash: ‘anti-AI/anti-tech’ sentiment as a rising axis
- •China strategy as commoditization (analogy: Google Docs undermining Office profits)
- •Copyright complexity: protects creators and historically enabled software business models
- 1:13:52 – 1:18:08
From reactive to proactive tech governance: model laws, jurisdictional competition, ‘Elon Salvador’
In closing, Balaji proposes competing with restrictive regulation by shopping jurisdictions: drafting model legislation, building a policy ‘sales team,’ and partnering with pro-tech states/countries to create innovation-friendly environments. The conversation ends on the idea of applying ‘antitrust’ logic to governments via jurisdictional choice rather than pleading within a single regulatory regime.
- •Write ideal policy playbooks and model bills across states/countries
- •Prioritize pro-tech jurisdictions and offer investment in exchange for enabling laws
- •Use competition among governments to reduce arbitrary constraints
- •Reframe: treat the state as the monopoly and create alternatives via jurisdictional choice
