The incredible exploding AI budget | PitchBook News Summary

1. Key Themes

Theme 1: AI Observability Is Emerging as a High-Growth VC Category

LLM observability — tools helping enterprises monitor, debug, and optimize AI spend — is becoming a standalone investment category within infrastructure SaaS, bucking broader VC slowdowns.

"Within infrastructure SaaS, developer operations startups continue to defy a broader slowdown in VC investing. Since 2023, deal value has risen to $6.2 billion in capital across 133 pacts, up 55% over 2024's $4 billion in venture dollars invested into 157 deals... And AI observability startups are leading the pack."

Theme 2: Enterprise AI Spend Is Exploding — and Becoming a New Line-Item Category

AI spending is tripling year-over-year and is being reframed by enterprise leaders not as infrastructure cost, but as a labor-equivalent expense that demands scrutiny.

"We're going to have to accept that [AI] is going to be a major item of cost in any organization, and it's to be compared to the cost of people, and not to be compared to the cost of TCP/IP packets or compute or anything like that." — Marco Argenti, Goldman Sachs CIO

Supporting data: Enterprise AI spending tripled from $11.5 billion in 2024 to $37 billion in 2025 (Menlo Ventures), and OpenAI reported that average reasoning token consumption per organization increased 320x from 2024 to 2025.

Theme 3: European PE Is Pivoting to Defensive Deal Structures Amid Geopolitical Shock

The war in Iran disrupted European dealmaking momentum in Q1 2026, pushing sponsors toward risk-minimizing strategies like add-ons and club deals.

"Add-ons' share of dealmaking might be a better indicator of current market conditions. As a percentage of buyout deal count, add-ons reached a decade-high of 71.4%, up from 67.4% in 2025."

"Club deals are also taking a larger share of buyout dealmaking. Their share of total deal value rose by nearly 10 percentage points from the 2025 average, to 43.3%."

Theme 4: Frontier AI Models Are Creating Both Opportunity and Disruption in Cybersecurity

Anthropic's new Mythos Preview model, capable of identifying cybersecurity vulnerabilities, is expected to reshape the competitive landscape for cybersecurity startups — helping some while threatening others.

"Anthropic's Mythos Preview model, with its abilities to identify cybersecurity vulnerabilities, is expected to be a boon for some cybersecurity startups and a competitive threat to others."

Theme 5: Record VC Quarter for Female-Founded Companies — Driven by Mega AI Rounds

Startups with at least one female founder hit a record $166 billion in Q1 2026 — but the headline number is heavily concentrated in a few massive AI raises.

"Startups with at least one female founder have had their best quarter ever, garnering more than $166 billion in the first three months of 2026. This was powered by massive rounds for AI companies like OpenAI and Anthropic."

2. Contrarian Perspectives

Cheaper Tokens Don't Reduce AI Bills — They Inflate Them

The conventional assumption is that falling token prices should lower enterprise AI costs. The data says the opposite: cheaper tokens unlock more use cases, which drives consumption up exponentially, resulting in higher overall spend.

"While token prices have fallen — down 300x for commodity models and down 12x for frontier models since, according to Tokencost — spending on AI has gone up. With cheaper tokens enabling more use cases and more complex tasks requiring more tokens, companies' bills have also gone up."

Supporting evidence: OpenAI reported a 320x increase in reasoning token consumption per organization from 2024 to 2025, even as per-token prices fell sharply.

European PE Mega-Deals Are a Lagging Indicator, Not a Signal of Current Confidence

The strong Q1 2026 mega-deal numbers (15 deals worth €44.2 billion) may tempt observers to read market confidence into European PE — but these deals were negotiated before the war in Iran broke out and reflect H2 2025 momentum, not current conditions.

"Given the size of the deals, they were almost certainly already in negotiation before the war broke out and hence reflect the strong momentum of H2 2025 rather than the current scene."

AI Is a People-Cost Substitute, Not a Tech-Cost Line Item

The prevailing framing treats AI spend as a compute or infrastructure cost to be optimized by procurement. Goldman Sachs' CIO argues it should be evaluated against headcount costs instead — a reframe with major implications for how CFOs allocate and justify AI budgets.

"It's to be compared to the cost of people, and not to be compared to the cost of TCP/IP packets or compute or anything like that." — Marco Argenti, Goldman Sachs CIO

3. Companies Identified

Anthropic

Description: AI safety and frontier model company
Why mentioned: Its Mythos Preview model can identify cybersecurity vulnerabilities, creating both competitive threats and opportunities for cybersecurity startups
Quote: "Anthropic's Mythos Preview model, with its abilities to identify cybersecurity vulnerabilities, is expected to be a boon for some cybersecurity startups and a competitive threat to others."

OpenAI

Description: Leading AI model provider
Why mentioned: Disclosed 320x increase in reasoning token consumption per enterprise organization from 2024 to 2025, illustrating the explosion of enterprise AI spend
Quote: "OpenAI disclosed in its most recent state of enterprise AI research that the average reasoning token consumption per organization increased 320x from 2024 to 2025."

AMP (VC Firm)

Description: VC firm founded by former Andreessen Horowitz partner Anjney Midha
Why mentioned: Raised $1.3 billion for its first-ever fund — a remarkable debut raise signaling strong LP conviction in the fund's AI-focused thesis
Quote: "AMP, a VC firm founded by former Andreessen Horowitz partner Anjney Midha, raised $1.3 billion for its first fund."

Aria Networks

Description: AI networking infrastructure startup
Why mentioned: Secured a $125 million Series A, one of the largest early-stage AI infrastructure deals in the period
Quote: "AI networking infrastructure startup Aria Networks secured a $125 million Series A."

Sidewinder Therapeutics

Description: Cancer-focused biopharma startup
Why mentioned: Raised a $137 million Series B led by Frazier Life Sciences and Novartis Venture Fund
Quote: "Cancer-focused biopharma startup Sidewinder Therapeutics raised a $137 million Series B."

Patlytics

Description: AI platform for patent workflows
Why mentioned: Raised a $40 million Series B led by SignalFire; represents the application of AI to legal/IP workflows
Quote: "Patlytics, a startup developing an AI platform for patent workflows, received a $40 million Series B led by SignalFire."

Jeito Capital

Description: French PE firm focused on biopharma
Why mentioned: Closed its second biopharma fund at $1.2 billion, betting on opportunity created by the $180 billion pharma patent cliff
Quote: "French PE firm Jeito Capital closed its second biopharma fund on $1.2 billion, as investors bet that the $180 billion pharma patent cliff will create more attractive opportunities in the space."

Polygon Labs

Description: Stablecoin-powered payments platform
Why mentioned: In talks to raise $50–$100 million, signaling continued investor interest in crypto payments infrastructure
Quote: "Polygon Labs, which is building a stablecoin-powered payments platform, is in talks to raise between $50 million and $100 million."

Life Biosciences

Description: Biotech startup focused on aging
Why mentioned: Raised an $80 million Series D, reflecting continued capital flow into longevity biotech
Quote: "Biotech startup Life Biosciences, which focuses on aging, raised an $80 million Series D."

Mubadala Capital

Description: Abu Dhabi sovereign wealth-backed investment firm
Why mentioned: Raised nearly $1 billion for its third fund targeting Brazil — notable sovereign capital deployment into emerging markets
Quote: "Mubadala Capital raised almost $1 billion for its third fund targeting investments in Brazil."

154 Partners

Description: Sports-focused PE firm
Why mentioned: Closed its debut fund on $400 million, reflecting institutional appetite for sports as an alternative asset class
Quote: "154 Partners, a sports-focused PE firm, closed its debut fund on $400 million."

Trent AI

Description: London-based agentic AI security startup
Why mentioned: Raised a $13 million seed round, signaling early-stage investment in securing AI agent deployments
Quote: "London-based agentic AI security startup Trent AI raised a $13 million seed investment led by LocalGlobe and Cambridge Innovation Capital."

Associated British Ports

Description: UK port terminal operator
Why mentioned: KKR, Global Infrastructure Partners, and Brookfield are competing for a majority stake — a major infrastructure PE deal in progress
Quote: "KKR, Global Infrastructure Partners and Brookfield Asset Management are among PE firms interested in acquiring a majority stake in UK port terminal operator Associated British Ports."

XpFibre

Description: France-based fiber optic company
Why mentioned: KKR, DigitalBridge, and Brookfield are in bidding for the company, potentially valued at ~€8 billion
Quote: "KKR, DigitalBridge Group and Brookfield Asset Management are among firms selected to enter the next round of bidding for France-based fiber optic company XpFibre, which could be valued at around €8 billion."

4. People Identified

Marco Argenti

Description: Chief Information Officer, Goldman Sachs
Why mentioned: Offered a reframing of AI spend as a people-cost equivalent rather than an infrastructure cost, with significant implications for enterprise budgeting
Quote: "We're going to have to accept that [AI] is going to be a major item of cost in any organization, and it's to be compared to the cost of people, and not to be compared to the cost of TCP/IP packets or compute or anything like that."

Anjney Midha

Description: Founder of AMP VC; former partner at Andreessen Horowitz
Why mentioned: Raised $1.3 billion for AMP's debut fund, a landmark first close for a new AI-focused firm spun out of a top-tier VC
Quote: "AMP, a VC firm founded by former Andreessen Horowitz partner Anjney Midha, raised $1.3 billion for its first fund."

Derek Hernandez

Description: Analyst, PitchBook
Why mentioned: Provided analytical framing for why AI observability is a durable investment category, not a transient trend
Quote: "Most companies have limited visibility into where their AI spend is going, which models are delivering value, and where they're burning tokens on low-impact tasks. That gap is driving demand for observability tooling, and it will accelerate as AI usage continues to scale."

5. Operating Insights

Build AI Spend Visibility Before You Scale Usage

The article makes clear that enterprise AI budgets are exploding without corresponding visibility into ROI. Companies deploying AI broadly without observability tooling risk token waste on low-impact tasks.

"Most companies have limited visibility into where their AI spend is going, which models are delivering value, and where they're burning tokens on low-impact tasks."

Tactical takeaway: Before expanding AI deployments, operators should implement LLM observability tools to track per-workflow token consumption, model performance, and cost-to-output ratios. This is increasingly table-stakes at scale.

In Uncertain Markets, Default to Add-Ons Over New Platforms

European PE sponsors are responding to geopolitical uncertainty by dramatically increasing the share of add-on deals (now a decade-high 71.4% of buyout deal count) rather than committing to new platform investments.

"Rather than committing capital to new platform investments, sponsors are managing their risks by adhering to their existing portfolios. Add-ons are also typically smaller and a lower-risk way to deliver growth through targeted M&A within the same vertical."

Tactical takeaway: In high-uncertainty environments, acquirers and operators should prioritize bolt-on acquisitions within known verticals over greenfield platform bets — preserving optionality while still deploying capital productively.

6. Overlooked Insights

The $180 Billion Pharma Patent Cliff Is Actively Driving Dedicated Fund Formation

While briefly mentioned in the context of Jeito Capital's fund close, the $180 billion pharma patent cliff is a structural market event large enough to support dedicated biopharma PE funds — and is attracting meaningful LP capital specifically because of it.

"French PE firm Jeito Capital closed its second biopharma fund on $1.2 billion, as investors bet that the $180 billion pharma patent cliff will create more attractive opportunities in the space."

This is a durable, calendar-driven tailwind for biopharma investors that is underappreciated relative to the AI coverage dominating VC attention.

2021 Vintage Real Assets Funds Have a Median DPI of Just 0.08x

The benchmark data quietly buried at the bottom of the newsletter reveals that 2021 vintage global real assets funds have returned only 8 cents per dollar invested to LPs (median DPI of 0.08x), despite a median IRR of 12.40%. This signals that returns are largely unrealized and distributions remain extremely low — a potential liquidity concern for LPs in this vintage.

"Median DPI: 0.08x" — Daily Benchmark: 2021 Vintage Global Real Assets Funds

For investors evaluating secondaries or continuation vehicles in this vintage, the gap between paper IRR and actual distributions is a critical due diligence flag.