Inside the inference AI ‘bull run’

1. Key Themes

Inference chips are having a record funding year, driven by agentic AI

Startups building inference chips have raised $4.18 billion across 22 deals through Aug. 28, putting them on track to beat last year's record.
- Quote: "Startups that build inference chips—the chips that AI models run on—are on track for a record year of VC funding. These businesses raised $4.18 billion across 22 deals through Aug. 28... and are on track to surpass last year's record of $5.21 billion, generated by 31 deals."
Agentic AI adoption is cited as the main demand driver.
- Quote: "The adoption of agentic AI—tools that can pursue their own goals, rather than just respond to prompts—is a huge driver of funding, said senior analyst and author of the report, Dimitri Zabelin."
Inference chips differ from GPUs by being purpose-built for serving models rather than training them, with a faster and cheaper value proposition.
- Quote: "Inference chips are different from GPUs in that they're specifically built to answer questions asked of an AI model. They are designed to be faster and cheaper than GPUs, which are more general-purpose chips used to train AI models."
Capital is concentrating into fewer, larger rounds. 2026 is on pace to top last year's total with far fewer deals (22 so far vs. 31 for all of last year).
- Quote: "raised $4.18 billion across 22 deals through Aug. 28... surpass last year's record of $5.21 billion, generated by 31 deals."

Secondary-market liquidity is stretched by duration risk and the exit drought

Secondaries have been a key liquidity source for LPs, with record volume.
- Quote: "Last year, secondary transaction volume reached a record $226 billion, up 41% from 2024, which was itself up more than 40% year over year, according to data from intermediary Evercore."
Secondary buyers now worry about how long assets take to exit and whether they are paid for that risk.
- Quote: "One of the biggest risks we take as secondary buyers when pricing and underwriting deals is duration risk... The key question is whether we are appropriately pricing that risk and being adequately compensated for taking it on."
Continuation-fund assets are being rolled into new continuation vehicles ("CV squared"), a sign of exit difficulty.
- Quote: "Instead of being sold at the end of their three- to five-year life, growing numbers of companies held in continuation funds are being transferred into new continuation funds, so-called CV squared transactions."

Mega-rounds continue across defense, physical AI, industrial software, and AI infrastructure

Large rounds at multibillion-dollar valuations span defense/drones, AI for critical infrastructure, and manufacturing software.
- Quote: "Netherlands-based drone maker Destinus is in talks to raise a new round at a €6 billion valuation, with peer Helsing investing in the round."
- Quote: "ASI, the developer of AI for the critical infrastructure and security sectors, secured a $325 million Series C led by Spark Capital at a $3 billion valuation."
- Quote: "Manufacturing software developer Arda Industries launched from stealth with $232 million: A $160 million Series A led by Altimeter, Accel and Khosla Ventures, and a $72 million seed led by Accel and Founders Fund."
Autonomy and physical-world AI are drawing large checks.
- Quote: "Neolix Technologies, which builds autonomous delivery vehicles, raised a $400 million Series E led by Boyu Capital."
- Quote: "Reliable Robotics added $115 million to its Series D, including a $25 million term loan from Silicon Valley Bank and bringing the round's total to $275 million."

Sports is emerging as an institutional investment category

Several sports-focused capital formations appeared in a single day.
- Quote: "The NFL's investment arm, 32 Equity, is partnering with General Atlantic to form a sports investing platform."
- Quote: "Clara Vista Investment Partners, one of the main backers of English Premier League club Ipswich Town FC, wants to raise $250 million to invest in more football teams, Bloomberg reported."

2. Contrarian Perspectives

Higher valuations may hurt returns in the hottest sector

Record funding and soaring valuations are not unambiguously good for investors. They raise the bar for exits and shrink the buyer universe, so strategic acquirers are becoming pickier even as prices rise.
- Evidence: "'higher valuations raise the exit price needed to deliver attractive investor returns,' Zabelin added."
- Evidence: "Higher valuations narrow the range of feasible acquirers, said Zabelin, who expects most exits in the space to be strategic acquisitions by bigger tech companies and established chipmakers. Still, the price has to be right, and buyers are getting more selective."

Secondaries may be a liquidity "lifeline" that is itself running out of liquidity

Record secondary volume masks a structural problem: buyers are accumulating assets that are taking longer to exit than anticipated. Continuation-fund rollovers may signal unsellable assets rather than prized holdings.
- Evidence: "While some of these businesses are top performers that the manager wants to hold on to indefinitely, investors suspect that many are businesses the manager wants to exit, but still can't find a buyer for."
- Evidence: "There are signs, however, that the assets hoovered up by secondary funds are taking longer to exit than the fund managers anticipated."

Weak IPO debut for an AI infrastructure company despite the AI boom

Public markets are not uniformly rewarding AI infrastructure, even as private valuations climb.
- Evidence: "Elice, an AI infrastructure company backed by GS Ventures, fell 32% on its debut on South Korea's stock exchange." The newsletter also notes "OpenAI and Anthropic's delayed IPOs" as part of the current VC uncertainty.

3. Companies Identified

Inference chip startups (category, unnamed)

Description: Startups building chips optimized for running AI models.
Why mentioned: Record VC funding; central subject of the PitchBook report "Inference Chips Are Just Starting Their Bull Run."
Quotes: "Startups that build inference chips—the chips that AI models run on—are on track for a record year of VC funding."

OpenAI

Description: AI lab.
Why mentioned: Acquired Orchestra Bio; its delayed IPO is cited as a source of VC market uncertainty.
Quotes: "OpenAI acquired Orchestra Bio, an 8VC-backed developer of a drug development coordination platform." Also: "OpenAI and Anthropic's delayed IPOs."

Anthropic

Description: AI lab.
Why mentioned: Delayed IPO cited as market uncertainty.
Quotes: "from OpenAI and Anthropic's delayed IPOs to rising rates, geopolitical pressure and more"

Destinus

Description: Netherlands-based drone maker.
Why mentioned: In talks to raise at a €6 billion valuation with Helsing investing.
Quotes: "Destinus is in talks to raise a new round at a €6 billion valuation, with peer Helsing investing in the round, Bloomberg reported."

Helsing

Description: Defense tech peer of Destinus.
Why mentioned: Investing in Destinus's round.
Quotes: "with peer Helsing investing in the round"

Neolix Technologies

Description: China-based autonomous delivery vehicle maker.
Why mentioned: $400M Series E led by Boyu Capital.
Quotes: "raised a $400 million Series E led by Boyu Capital."

ASI

Description: Developer of AI for critical infrastructure and security.
Why mentioned: $325M Series C at a $3B valuation led by Spark Capital.
Quotes: "secured a $325 million Series C led by Spark Capital at a $3 billion valuation."

Otter.ai

Description: AI transcription and note-taking platform.
Why mentioned: Raised $320M from General Catalyst's Customer Value Fund.
Quotes: "raised $320 million from General Catalyst's Customer Value Fund."

Arda Industries

Description: Manufacturing software developer.
Why mentioned: Launched from stealth with $232M across a Series A and seed.
Quotes: "launched from stealth with $232 million: A $160 million Series A led by Altimeter, Accel and Khosla Ventures, and a $72 million seed led by Accel and Founders Fund."

Reliable Robotics

Description: Autonomous aircraft developer.
Why mentioned: Added $115M to Series D, including a $25M term loan.
Quotes: "added $115 million to its Series D, including a $25 million term loan from Silicon Valley Bank and bringing the round's total to $275 million."

Blackwall

Description: Estonia-based web traffic management tools developer.
Why mentioned: $100M Series C at a $2.2B valuation.
Quotes: "raised a $100 million Series C led by AVP at a $2.2 billion valuation."

SendCutSend

Description: Custom metal parts provider.
Why mentioned: Series B at a $2.1B valuation.
Quotes: "raised a Series B led by Spark Capital at a $2.1 billion valuation."

Different

Description: B2B marketing system developer.
Why mentioned: Emerged from stealth with a $48M seed.
Quotes: "emerged from stealth with a $48 million seed round led by General Catalyst."

Daytona

Description: Agentic cloud startup.
Why mentioned: $60M Series B at a $1B valuation.
Quotes: "raised a $60 million Series B led by FirstMark at a $1 billion valuation."

Qualitate

Description: Enterprise primary research system developer.
Why mentioned: $25M Series A.
Quotes: "secured a $25 million Series A led by Insight Partners."

AuxoAI

Description: Enterprise AI developer.
Why mentioned: $20.7M Series A.
Quotes: "raised a $20.7 million Series A led by NTTVC."

Firefly

Description: Physical AI safety platform developer.
Why mentioned: Emerged from stealth with a $20M Series A.
Quotes: "emerged from stealth with a $20 million Series A led by Definition Capital."

IntraFi

Description: Deposit network for banks.
Why mentioned: Blackstone and Warburg Pincus considering a sale at over $10B.
Quotes: "which operates a deposit network for banks and could be valued at over $10 billion in a deal, Reuters reported."

Mynt (GCash operator)

Description: MUFG-backed payments platform operator.
Why mentioned: Raised ~$967M in its Philippines IPO.
Quotes: "raised about $967 million in its Philippines IPO."

Orchestra Bio

Description: 8VC-backed drug development coordination platform.
Why mentioned: Acquired by OpenAI.
Quotes: "OpenAI acquired Orchestra Bio, an 8VC-backed developer of a drug development coordination platform."

Elice

Description: AI infrastructure company backed by GS Ventures.
Why mentioned: Fell 32% on its South Korea debut.
Quotes: "fell 32% on its debut on South Korea's stock exchange."

Nvidia

Description: Chip and AI infrastructure company.
Why mentioned: Investing in four German startups.
Quotes: "announced investments in four German startups: Agile Robots, Avelios Medical, Mirelo and Reflex Aerospace."

Muse (Meta)

Description: Meta product with share features.
Why mentioned: Security issue exposed personal data via shared links.
Quotes: "Thousands of links were found to contain personal information like loan applications, bank fraud alerts and internal performance reports. Meta suspended sharing yesterday in response."

Ara Partners

Description: Decarbonization-focused investor.
Why mentioned: Closed Ara Energy Fund I.
Quotes: "closed on $2.5 billion for Ara Energy Fund I."

Monomoy Capital Partners

Description: PE firm.
Why mentioned: Closed inaugural family fund.
Quotes: "closed on $589 million for its inaugural Monomoy Family Fund."

32 Equity / General Atlantic

Description: NFL investment arm and growth investor.
Why mentioned: Forming a sports investing platform.
Quotes: "The NFL's investment arm, 32 Equity, is partnering with General Atlantic to form a sports investing platform."

Clara Vista Investment Partners

Description: New York-based backer of Ipswich Town FC.
Why mentioned: Seeking $250M for more football teams.
Quotes: "wants to raise $250 million to invest in more football teams, Bloomberg reported."

Blue Owl / Ufenau Capital Partners

Description: Alternative asset manager and Swiss PE firm.
Why mentioned: Blue Owl took a passive minority stake.
Quotes: "Blue Owl acquired a passive minority stake in Ufenau Capital Partners, a Swiss PE firm, The Drawdown reported."

Tikehau Capital

Description: Paris-based asset manager.
Why mentioned: Seeking to sell a stake to a strategic investor.
Quotes: "is looking to sell a stake to a new strategic investor, Bloomberg reported."

Evercore

Description: Secondary-market intermediary.
Why mentioned: Source of record secondary volume data.
Quotes: "according to data from intermediary Evercore."

4. People Identified

Dimitri Zabelin

Description: Senior analyst at PitchBook and author of the inference chips report.
Why mentioned: Source of the key analysis on inference chip funding, valuations, and exits.
Quotes: "The adoption of agentic AI... is a huge driver of funding, said senior analyst and author of the report, Dimitri Zabelin." Also: "'higher valuations raise the exit price needed to deliver attractive investor returns,' Zabelin added."

Unnamed partner at a large secondaries shop

Description: Secondary-market investor.
Why mentioned: Articulated duration risk as the central underwriting challenge.
Quotes: "One of the biggest risks we take as secondary buyers when pricing and underwriting deals is duration risk."

5. Operating Insights

Build exit paths into your capital strategy for deep-tech and hardware

For capital-intensive startups like inference chip makers, plan for strategic acquisition as the likely exit, and be mindful that inflated valuations narrow your potential buyers.
- Quote: "Higher valuations narrow the range of feasible acquirers, said Zabelin, who expects most exits in the space to be strategic acquisitions by bigger tech companies and established chipmakers."

Position products around agentic workloads and inference economics

Demand for faster, cheaper inference is tied to agentic AI adoption, so products that reduce cost-per-query for autonomous agents align with where funding is flowing.
- Quote: "They are designed to be faster and cheaper than GPUs, which are more general-purpose chips used to train AI models."

Treat shareable links and AI outputs as a security surface

The Muse incident shows that convenience features like share links can leak sensitive data at scale. Audit sharing defaults, access controls, and what gets indexed or exposed.
- Quote: "Thousands of links were found to contain personal information like loan applications, bank fraud alerts and internal performance reports."

6. Overlooked Insights

Non-dilutive and structured capital is creeping into late-stage rounds

Reliable Robotics' Series D includes a bank term loan, and Otter.ai raised from a customer-value-oriented fund rather than a classic equity round. This points to blended capital structures for scaling companies.
- Quote: "including a $25 million term loan from Silicon Valley Bank and bringing the round's total to $275 million."
- Quote: "raised $320 million from General Catalyst's Customer Value Fund."

ESG engagement remains broad despite headline backlash

A PitchBook survey shows the large majority of respondents still engage with sustainable investing, and only a small share have disengaged, mostly GPs and companies.
- Quote: "80% of 193 respondents indicated they engage with ESG or support the sustainable investment industry. 4% reported they once engaged with ESG but no longer do so. Of the 4%, most are GPs and companies."