04-09-2026Price:

The Frontier

Your signal. Your price.

AI & TECH

Aaron Levie warns AI agents will collapse enterprise software valuations

Thursday, April 9, 2026 · from 2 podcasts
The a16z ShowForward Guidance
  • AI agents will replace seat-based SaaS models, making traditional cash flow projections impossible.
  • Security risks and legacy systems will slow enterprise adoption, creating a startup advantage.
  • R&D costs are shifting from fixed salaries to volatile token budgets, upending corporate finance.

The traditional software company is becoming un-investable.

Box CEO Aaron Levie argues that the coming wave of autonomous AI agents will obliterate the seat-based licensing model that built the SaaS industry. On *The a16z Show*, he said software must be redesigned for a future where agents outnumber human users by “three orders of magnitude.” Its value will shift from a polished UI to durable APIs agents can reliably navigate. The certainty of growth that supported a decade of high S&P 500 valuations is vanishing.

“The era of valuing software companies via Discounted Cash Flow is over. In a world of exponential disruption, no analyst can reliably project a company’s cash flows three years out.”

- Jordi Visser, Forward Guidance

Jordi Visser, on *Forward Guidance*, agrees. He argues software moats are evaporating because AI agents make the human-centric model obsolete. This creates a valuation death trap. He predicts a massive wealth transfer from traditional equities as investors realize software moats are “actually sieves.”

Yet the transition won’t be swift. Levie notes that while a startup can deploy agents with total context, a bank like JPMorgan faces existential security risks from prompt injection or a “rogue agent” leaking M&A data. This will delay the “write” phase of enterprise AI for years, creating a massive agility gap. Steve Sinofsky, former Microsoft executive, predicts a widening adoption chasm between nimble startups and legacy-constrained giants.

“We are moving from a world where R&D costs are dominated by human salaries to one where ‘engineering compute budgets’ - the cost of tokens - become a volatile, mission-critical line item.”

- Martin Casado, The a16z Show

The financial model of building software is also fracturing. As Martin Casado observes, infrastructure spend across his portfolio is going “asymptotic” because AI enables a massive increase in total software production. CFOs now face a new debate: what percentage of R&D spend, typically 14-30% of tech company revenue, should go to token consumption instead of human payroll? This shift from fixed-cost engineering to elastic, volatile compute budgets creates a FinOps nightmare, where a single inefficient algorithm could swallow quarterly earnings.

Visser frames this as part of a broader compute supercycle, driven by the shift from the “chatbot era” to the “Agentic era” in late 2023. He notes this requires a 1,000x increase in compute power, creating a vertical demand curve for hardware that markets haven’t yet digested.

The consensus is that software valuations are entering a period of radical uncertainty. The organizations that win will be those that stop trying to cap AI spend and instead learn to manage the output of a thousand-agent workforce.

AgentsEnterpriseMarkets

By the Numbers

  • 100agent-to-human ratiometric
  • 1000agent-to-human ratiometric
  • 50visible portfolio companiesmetric
  • 6months of growthmetric
  • 14%R&D as percentage of revenue (low end)metric
  • 30%R&D as percentage of revenue (high end)metric

Entities Mentioned

CasaCompany
JP Morganinstitution
MicronCompany
OpenAItrending
Opusmodel
PerplexityCompany
Wall StreetConcept

Source Intelligence

What each podcast actually said

The a16z Show
The a16z Show

The Agent Era: Building Software Beyond Chat with Box CEO Aaron Levie • Apr 8

  • Aaron Levie argues that the diffusion of AI capability across enterprises will be slower than Silicon Valley expects, citing entrenched domain knowledge in systems like SAP and new security and operational complexities.
  • The central enterprise question is how to build software for a future where AI agents outnumber human users by factors of 100 or 1000 to one. This shifts focus to designing robust APIs, access controls, and monetization for agents.
  • A successful emerging paradigm gives coding agents access to SaaS tools and internal workflows, enabling them to both read information and use APIs or write code to execute tasks. This is exemplified by tools like OpenAI's 'super app' and Perplexity Computer.
  • Steve Sinofsky observes that agents do not seek simpler interfaces but choose backends based on cost, durability, and reliability. He contends the industry's focus on marketing to agents via APIs is wrong, as agents select systems based on underlying quality, not interface polish.
  • A major operational challenge is coordinating thousands of autonomous agents acting on shared systems, like a Box repository, which risks creating conflicting operations, performance issues, and security vulnerabilities that CFOs and CIOs must manage.
  • The permission model for agents is complex. While the 'end-to-end argument' suggests treating them like separate humans with their own accounts, agents are legally extensions of their users, requiring full oversight and lacking a right to privacy, which breaks traditional RBAC models.
  • Current AI agents struggle with information containment, as data in the context window can potentially be extracted via prompt injection. This makes it difficult to securely grant agents access to highly confidential resources like M&A data rooms.
  • Sinofsky predicts a widening gap in adoption speed between startups, which can adopt agents freely, and large enterprises like JP Morgan, which face significant legacy system and risk constraints, slowing AI diffusion.
  • There is tension between legacy SaaS vendors and the agent ecosystem, as agents want unlimited API access to data for operations, while vendors have traditionally monetized intelligence and domain expertise through UI-based subscriptions, not pure data licensing.
  • The engineering compute budget for AI tokens is becoming a critical financial debate. CFOs must decide what percentage of R&D spend should go to tokens, a decision that directly impacts earnings per share given R&D typically constitutes 14% to 30% of tech company revenue.
  • Sinofsky argues Wall Street is mis-modeling the AI economic opportunity by assuming a fixed revenue pie. He draws parallels to the PC and cloud eras, where new usage models created demand orders of magnitude larger than initially projected.

Also from this episode:

AI & Tech (3)
  • Martin Casado notes that every infrastructure company in his portfolio of about 50 has seen asymptotic growth in the last six months due to an unprecedented increase in software being written, driven by AI agent development.
  • A key friction is the current high cost of tokens, which pushes the industry toward usage-based pricing. This creates a short-term budgeting nightmare for engineering teams deciding between experimental waste and perfect optimization.
  • Sinofsky contends the token cost issue is transitional, comparing it to historical transitions like mainframe MIPS pricing. He believes the cost will plummet due to increased supply, algorithmic improvements, or hardware changes, making compute abundant.
AgentsEnterpriseAI InfrastructureStartups
Forward Guidance
Forward Guidance

Why AI Will Reprice The Entire Economy | Jordi Visser • Apr 6

  • Jordi Visser argues we entered the Agentic era in late November, driven by releases like Opus 4.5, where compute demand is already a thousand times higher than the chatbot era.
  • Visser says software companies can no longer be valued with discounted cash flow models because AI progress is too disruptive, which makes Bitcoin an attractive growth asset without cash flows.
  • Visser prefers silver over gold and semiconductors as hardware plays, noting silver is up 60% in six months and is a critical component in drones and technology.
  • He distinguishes Mag7 hardware companies like Nvidia, Tesla, and Apple from software companies, calling Microsoft a disaster and noting Micron trades at a forward P/E below 4.

Also from this episode:

Business (2)
  • Visser predicts CPI will exceed 4% year-over-year for the next two months, creating a period to unwind positions before a recession narrative presents a buying opportunity for stocks.
  • He contends the S&P 500 rose 15% and U.S. household net worth increased $15 trillion over the last year, making oil price shocks less relevant to an economy now driven by AI spending.
AI & Tech (4)
  • The labor arbitrage from AI favors solo entrepreneurs over enterprises, Visser says, as his annual cost for five LLMs and hardware is $17,000, far cheaper than human employees.
  • Visser argues AI will not cause mass unemployment due to a domestic labor shortage and demographic issues, but will destroy the corporate ladder, creating psychological damage in the job market.
  • He views AI as a nuclear weapon for militaries and an existential spend for big tech, forecasting a murky future where government control could compress multiples for private AI companies.
  • Visser recommends building a relationship with AI through verbal conversation as a primary learning method, suggesting daily use is essential to gain proficiency.
MacroBankingMarketsBTC MarketsAgents

Related Stories