"the first is somewhat obvious but very important which is it's very hard to disrupt a market to break into a market unless there's something macro that's happening that's an enormous change obviously AI is that" - Mike Volpi [00:01:20]
"you have to be very careful about your past success... the more success a firm has had the more to use an AI term reinforcement learning there is of how things were done" - Mike Volpi [00:02:13]
Disclaimer: Orignal content owned by or sourced from third parties. It does not represent the views of 'Nuggets' platform or it's team. AI is used extensively across this platform including for summaries. Accuracy is not guaranteed, there can be mistakes. Any info or content on this platform is not a financial, legal, or investment advice. Do your own research. Refer for complete disclosures:- Terms of Use · Full Disclaimer
"open source is a is a relevant phenomenon but not a business in AI if you look at the overwhelming amount of prompting that's happening it's against the most advanced models and that's the most monetizable" - Mike Volpi [00:23:42]
"we live in a world where Nvidia is the gatekeeper to all compute... I don't see that sustaining and I do see more specialization in the kind of compute" - Mike Volpi [00:34:50]
"I think the ultimate expression of self-confidence is when you can accept the fact that you don't know something." - Mike Volpi [00:55:13]
Speakers & Credentials
Jack Altman (Host): General Partner at Uncapped, serial entrepreneur, and host exploring the shifting dynamics of venture capital and technology.
Mike Volpi (Guest): General Partner and Co-Founder of Hanabi Capital (a new $175M AI-focused VC fund), former long-time General Partner at Index Ventures, and early operator at Cisco during its hyper-growth phase (scaling from 300 to 55,000 employees).
The Structural Death of Legacy VC: Mike Volpi asserts that the venture frameworks established during the 2010s SaaS boom are now toxic liabilities; past success operates as flawed "reinforcement learning" in an AI era where software creation costs have plummeted to zero.
The Foundational Monopoly: The core LLM war is fundamentally over, won by the "Big Five" (OpenAI, Anthropic, Google, Meta, xAI) due to an insurmountable capital and compute moat where incumbents spend $50B–$100B annually on chips.
The Neo-Lab Reality Check: Funding new general-purpose AI labs is financially irrational; the only viable "neo-labs" are those capturing proprietary, non-internet data, such as robotics companies pre-training on physical embodiment data.
The Compute Pivot: Nvidia's total dominance is temporary. The massive surge in inference demand is shifting the hardware market toward specialized, mission-specific ASICs (like Cerebras and Etched) that sacrifice flexibility for blistering speed.
The "Forward Deployed" Application Layer: Traditional seat-based SaaS is dying. The new high-margin enterprise model looks like Palantir—leveraging Forward Deployed Engineers (FDEs) to translate raw AI capability into massive, multi-million dollar business solutions.
The Beginner's Mind: The democratization of internet knowledge has accelerated founder maturity; today's 21-year-old founders possess the commercial acumen of previous generations' 35-year-olds, requiring veteran investors to completely strip their egos to remain competitive.
[00:00:00] - The Genesis of a New VC Firm (Hanabi Capital)
[00:02:55] - How AI Breaks the Traditional SaaS Economic Model
[00:05:53] - Stage Agnosticism & The Irrelevance of Legacy Deal Structures
[00:10:41] - The Evolution of Brand Building in Venture Capital
[00:20:50] - The Foundational Model Wars: Why the Big 5 Have Already Won
[00:27:12] - The Neo-Lab Fallacy & The Robotics Data Moat
[00:32:50] - The Compute Crunch and the Future of Silicon
[00:36:55] - AI Application Layer & The FDE (Forward Deployed Engineer) Model
[00:43:24] - Legacy SaaS Stocks & The "Elon Transition" Imperative
[00:45:51] - Geopolitics and the Pivot to Defense Tech
[00:47:51] - The Operator's Edge & Cisco's Historical Hypergrowth
[00:51:01] - Founder Evolution & Embracing the Beginner's Mind
3. Detailed Thematic Summary
The AI Macro Wave & The Destruction of Legacy VC Mechanics [00:01:20]
Breaking into the VC market requires a massive macroeconomic disruption. AI serves as this wedge, completely obsoleting the traditional SaaS business model [00:01:20].
Volpi introduces the concept of the "Past Success Trap," noting that VCs suffer from negative "reinforcement learning"—the highly successful strategies of the 2010s cloud era actively lead to bad investments in the low-marginal-cost AI era [00:02:13].
Legacy SaaS relied on a high fixed cost to build the first unit of software, followed by near-zero distribution costs. AI has driven the initial creation cost of software down to zero, meaning every assumption regarding GTM, pricing, and engineering must be blown up [00:03:25].
Volpi structured his new fund, Hanabi Capital ($175M AUM), to be completely stage-agnostic, ignoring legacy VC mandates of demanding 15% to 20% ownership at the Series A (which typically ranges from $15M to $30M today) [00:14:15].
He argues that investing in a dominant growth-stage AI company at a $60B valuation (like Anthropic) can legitimately yield 10x+ venture-scale returns, completely invalidating the old rule that growth investing only yields 3x multiples [00:07:10].
The era of bloat is over: Volpi mocks the 2021-era dynamic of having 600 people at a VC firm just to "do spreadsheets," arguing the future belongs to lean, full-stack individual investors [00:18:11].
Brand Building Nuance: Volpi argues against loud, banner-heavy VC marketing. True prestige must be unspoken—"if you know, you know"—comparing ideal VC branding to Mercedes-Benz [00:10:17] or Hermes [00:12:30]. He and Jack note that Thrive Capital executes this perfectly, ranking highly among elite college founders despite a quiet public footprint [00:11:53].
The Foundational Layer Monopoly & The Neo-Lab Squeeze [00:22:00]
The foundational LLM war is over. The "Big Five" (OpenAI, Anthropic, Google, Meta, and xAI) have insurmountable leads built on an impenetrable moat of capital and compute [00:22:00].
These incumbents are conservatively spending $50B to $100B per year on compute. Because of this, a new AI startup raising $2B is orders of magnitude behind and mathematically cannot compete on generalized intelligence [00:22:30].
While open source is a "relevant phenomenon," it is not a viable frontier business model. Monetization happens at the bleeding edge of prompting [00:23:42]. Furthermore, training frontier open source models has become so exorbitantly expensive that major players are quietly closing their newest models (e.g., the latest Qwen and Mistral models are now closed) [00:24:54].
Giving away $50B models out of the "kindness of your heart"—like Meta did initially with Llama, or like DeepSeek—is financially unsustainable over the long term [00:26:30].
The only justifiable "neo-lab" investments are those attacking proprietary data not available on the internet. Volpi highlights Periodic Labs [00:28:48] and Mind Robotics (a Rivian spin-out) [00:32:38], noting that physical embodiment pre-training data must be generated in the real world. Tesla's Optimus is also positioned to win here [00:32:28].
He explicitly warns against relying on data labelers like Scale AI, Surge AI, MTurk, or Turing as a defensible moat, arguing that basic annotation is a low-barrier, lowest-bidder commodity [00:30:25].
The Future of Silicon & The Post-Nvidia Hardware Landscape [00:33:28]
The hardware industry is facing a dual-load crisis: unprecedented massive training runs are running simultaneously alongside an explosion in inference demand, pushing GPU requirements purely vertical [00:33:28].
Global compute supply is rigidly bottlenecked by physical TSMC wafer starts. This reality heavily rewarded OpenAI, who brilliantly stockpiled compute ahead of the market [00:33:52].
Volpi predicts the imminent fracture of Nvidia's absolute hegemony. The future of compute relies on mission-specific silicon rather than general-purpose GPUs [00:34:50].
He frequently debates Cerebras CEO Andrew Feldman about the viability of their chips for training, but unequivocally champions Cerebras for its overwhelming advantages in pure inference speed [00:35:23].
The industry is moving toward ASICs: companies like Etched and Talis are evolving hardware further by hardcoding neural network weights directly into the silicon, removing flexibility to gain raw speed [00:35:49].
Domestic fab production is vital; Cerebras chips are currently manufactured in Arizona, and Intel is actively trying to bring more fabs online to reduce geopolitical reliance on Taiwan [00:36:34].
Application Layer Economics & The "Forward Deployed" Paradigm [00:36:55]
The historic VC disdain for service-heavy software businesses is dead. Ten years ago, VCs automatically passed on startups with "professional services" revenue; today, mimicking Palantir's model, Forward Deployed Engineers (FDEs) are the most valuable asset in enterprise software [00:41:43].
FDEs operate as critical translators between raw neural capabilities and highly bespoke corporate workflows. Because AI solves monumental business problems—like reducing T-Mobile's churn by 2%—startups can command massive contracts where the human labor cost of the FDE is a negligible rounding error [00:42:45].
Pre-AI legacy SaaS companies face an existential crisis. Standard SaaS stocks (like Workday) are trading at compressed multiples, but Volpi predicts visionary leaders like Dylan Field at Figma will figure out the transition [00:44:26].
Legacy companies must execute a visionary "Elon Transition"—convincing the market they are an AI/robotics company rather than just a car or software company—to survive [00:45:21].
Jack Altman notes a conversation with Brett (Berson) regarding systems of record: Volpi agrees that UIs won't entirely vanish, pointing out that Salesforce endures specifically because human interaction and muscle memory remain an incredibly high moat against pure AI agents [00:38:51].
Historical Context: Geopolitics, Operator Hypergrowth, & The Ego Check [00:45:51]
Geopolitics has fundamentally shifted the VC Overton Window. Volpi details how he initially rejected Anduril based on traditional Silicon Valley anti-weapons dogmas, before Founders Fund's Trae Stephens reframed weapons as the ultimate violence-deterrent system [00:47:17]. Europe's massive, permanent increase in defense spending validates this thesis [00:46:25].
Looking back at his exact 13-year operator history, Volpi notes he joined Cisco when it was 300 employees and rode it to 55,000 employees as they built the physical internet [00:48:22]. He explicitly references Steve Jobs' philosophy of "connecting the dots looking back" to explain how that historical hyper-growth experience gives him an edge today [00:49:06].
However, Volpi clarifies that operator experience isn't strictly mandatory to be a legendary VC. He cites Peter Fenton (whose biggest operation was a "lemonade stand"), former journalist Michael Moritz, and former sales rep John Doerr as proof that having a unique "angle" is more important than a traditional resume [00:49:55].
Finally, Volpi warns older investors that the internet has radically accelerated founder maturity. A 21-year-old founder today possesses the technical and commercial maturity of a 35-year-old from 15 years ago [00:51:16]. Veteran VCs must shed their egos and adopt a "beginner's mind" to treat these young founders as intellectual equals [00:53:10].
The Reference Vault
4. Data & Figures
Data Point
Value
Context
Timestamp
Hanabi Capital Fund Size
$175 million
Total AUM of Mike Volpi's new AI-focused boutique VC fund
The "Past Success" Reinforcement Trap [00:02:21]
Volpi masterfully adapts the AI concept of "reinforcement learning" to human organizational behavior. In stable business eras, operational success reinforces deeply held dogmas (like SaaS LTV/CAC ratios or strict gross margin rules). However, when the macroeconomic physics change—specifically, AI collapsing the marginal cost of software creation to zero—that very same "reinforced" behavior becomes toxic. VCs and operators trained in the 2010s are actively penalized by their own muscle memory, mistaking outdated SaaS heuristics for universal business truths.
The FDE (Forward Deployed Engineer) Arbitrage [00:41:43]
Historically, venture capital despised service-heavy businesses, demanding pure, zero-marginal-cost software margins. Volpi illustrates how AI violently flips this framework. Because AI systems target massive, bespoke corporate pain points (e.g., moving corporate churn by 2%), the traditional one-size-fits-all API SaaS model fails. Instead, deploying elite engineers (FDEs) to act as custom translators between raw neural capabilities and bespoke corporate workflows unlocks gargantuan, multi-million dollar contracts. At that enterprise scale, the human labor cost (COGS) of the FDE becomes an irrelevant rounding error.
The "Elon Transition" for Legacy Companies [00:45:21]
Volpi uses Tesla as the ultimate framework for legacy valuation survival. Tesla does not trade at car-company multiples because the market believes Elon Musk will successfully transition the company into a robotics and AI powerhouse. Similarly, legacy pre-AI SaaS companies (like Workday or Figma) are currently trading at depressed multiples (the "car company" equivalent). Their only path to survival is executing an "Elon Transition," proving to public markets that they can fundamentally reinvent their core architecture around native AI workflows before they bleed out.
The Beginner's Mind Imperative [00:53:10]
As high-signal knowledge distribution approaches zero cost via the internet and AI, the historic age-to-experience correlation has permanently broken. A 21-year-old founder today holds the commercial strategy and technical depth of a 35-year-old from the previous generation. Consequently, veteran investors must actively dismantle their ego-driven "all-knowing grandfather" persona. True institutional confidence is measured not by asserting past knowledge, but by the psychological flexibility to admit ignorance and engage 22-year-olds as absolute intellectual equals.
Mission-Specific vs. General Purpose Silicon [00:35:49]
The current computing era is defined by the "Nvidia Tax"—relying on general-purpose GPUs to handle both massive training runs and sprawling inference queries. Volpi predicts an inevitable hardware bifurcation. Just as crypto mining moved from generalized CPUs to hardcoded ASICs, AI compute will splinter. Next-generation hardware like Cerebras or Etched will hardcode neural weights directly into the silicon, explicitly abandoning flexibility in exchange for blistering, unmatched inference speed.
6. Anecdotes
Trae Stephens and the Morality of Deterrence [00:47:17]
While serving at Index Ventures, Volpi passed on investing in early defense-tech darling Anduril, bound by traditional Silicon Valley ESG "no weapons" dogmas. Trae Stephens from Founders Fund confronted Volpi, explaining the historical reality that weapons are fundamentally a deterrent system, and that establishing overwhelming capability is what actually prevents war. Volpi shares this anecdote to illustrate his total philosophical pivot toward defense tech investing, mapping the broader awakening of the venture capital industry to modern geopolitical realities.
The 100k Router vs. The T-Mobile Churn Reality [00:42:45]
To contextualize why enterprise software pricing models are undergoing a massive transformation, Volpi reminisces about his early career peddling routers at Cisco. Back then, a sales rep could justify charging an extra $100,000 merely because their hardware switch had 24 ports instead of 16. Today, AI isn't selling ports or API calls; it is promising the CEO of T-Mobile a 2% reduction in total customer churn. Volpi uses this story to hammer home the point: when you actually solve a multi-billion dollar business problem, the cost of custom engineering services (FDEs) is easily absorbed by the sheer gravity of the ROI.
The Cisco Hypergrowth Crucible & Steve Jobs' Advice [00:48:22]
Reflecting on his career, Volpi recalls joining Cisco as a tiny 300-person startup and riding the rocket ship to a 55,000-employee behemoth over 13 years. He channels Steve Jobs' famous "connect the dots looking back" philosophy to explain why this matters. The feeling of being at Cisco—literally manufacturing the routers that birthed the commercial internet—is the exact same momentum currently present in the foundational AI build-out. It serves as proof that he knows what a once-in-a-century societal infrastructure boom actually feels like from the inside.
Peter Fenton's Lemonade Stand [00:49:55]
To balance his own reliance on his operational past, Volpi tells a brief story about Benchmark's Peter Fenton. Despite Fenton being universally recognized as one of the greatest venture capitalists of all time, Volpi jokes that the biggest operating business Fenton ever ran was a "lemonade stand." He uses this anecdote (alongside mentions of journalist Michael Moritz and sales rep John Doerr) to prove to young investors that while having an "angle" is mandatory, having a strict CEO/operator resume is not.
7. References & Recommendations
Geopolitical & Financial Institutions
TSMC (Taiwan Semiconductor Manufacturing Company): Cited as the fundamental, physical bottleneck in global AI compute supply due to absolute limits on wafer starts. [00:33:52]
Companies & Firms
OpenAI: Recognized as a Big 5 winner; specifically praised for brilliantly stockpiling compute ahead of the market crunch. [00:22:00]
Anthropic: A Big 5 winner used to demonstrate that growth-stage entry valuations ($60B) can still return 10x in this new paradigm. [00:07:10]
Meta: A Big 5 winner currently burning billions on infrastructure to wage an open-source war against competitors. [00:22:00]
xAI: Included in the Big 5; Volpi believes their access to proprietary data loops secures their dominance, pending execution. [00:22:00]
Google: A Big 5 winner noted for holding a massive structural advantage via its proprietary TPU ecosystem. [00:35:09]
Hanabi Capital: Mike Volpi's new $175M boutique VC fund, built entirely to capitalize on the AI wave without legacy constraints. [00:14:15]
Cerebras: The hardware company Volpi strongly champions for pioneering inference-optimized silicon designed to break Nvidia's monopoly. [00:35:16]
Etched / Talis: Emerging chip manufacturers taking compute to the extreme by baking neural network weights directly into hardware ASICs. [00:35:49]
Intel: Referenced as attempting to bring massive domestic chip fabs online to ease global supply crunches. [00:36:34]
Palantir: Referenced as the godfather of the Forward Deployed Engineer (FDE) business model that modern AI enterprise startups are copying. [00:41:43]
Anduril: The defense tech prime that single-handedly forced Silicon Valley to abandon its aversion to military investing. [00:46:31]
Periodic Labs: A robotics startup backed by Volpi, highlighted for successfully generating highly proprietary, non-internet data. [00:28:48]
Mind Robotics: A Rivian spin-out backed by Volpi, highlighted for capturing a true AI moat by gathering data inside their own factories. [00:32:38]
Rivian: The electric vehicle manufacturer from which Mind Robotics successfully spun out. [00:32:38]
Tesla: The ultimate example of legacy valuation transformation via Elon Musk's pivot to robotics (Optimus) and AI. [00:45:21]
Scale AI / Surge AI / MTurk / Turing: A cluster of data labeling companies Volpi warns against relying on, as basic annotation is a low-barrier commodity. [00:30:25]
Thrive Capital: Praised as the gold standard for modern venture brand building—quiet, organic, and highly respected by young founders. [00:11:53]
Mercedes-Benz / Hermes: Used as analogies for how elite VC branding should function: unspoken prestige over loud advertising. [00:10:17]
Salesforce: The prime example of a legacy UI moat; humans are reluctant to abandon it because muscle memory is deeply entrenched. [00:38:51]
Figma / Workday: Juxtaposed to explain the fate of legacy SaaS. Figma has the visionary leadership to survive the AI pivot; Workday faces compression. [00:44:26]
T-Mobile: Used as the hypothetical enterprise client seeking a multi-billion dollar 2% churn reduction from an AI vendor. [00:42:45]
Cisco: Volpi's former employer, used as the historical benchmark for what a civilization-altering infrastructure boom looks like. [00:48:22]
People
Dario Amodei (Anthropic) & Sam Altman (OpenAI): Referenced to explain the critical importance of deep relationship capital; young VCs cannot easily access these veteran founders. [00:08:40]
Jensen Huang: Nvidia CEO, hypothetically referenced as an incumbent possessing the sheer capital to arbitrarily drop $10B on an open-source model purely to disrupt the market. [00:25:28]
Andrew Feldman: CEO of Cerebras, whom Volpi notes he frequently debates regarding the viability of Cerebras chips for LLM training. [00:35:23]
Trae Stephens: Founders Fund Partner who intellectually cornered Volpi into understanding the moral and geopolitical necessity of defense tech. [00:47:17]
Dylan Field: CEO of Figma, cited as the archetype of a visionary leader capable of successfully pivoting a traditional SaaS company through the AI epoch. [00:44:33]
Peter Fenton: Benchmark VC hailed as an all-time great investor, utilized to prove that prior operating experience is not a strict prerequisite for VC greatness. [00:49:55]
Michael Moritz: Legendary Sequoia VC referenced to show that even a former journalist can dominate venture capital. [00:50:11]
John Doerr: Legendary Kleiner Perkins VC referenced to show that a traditional sales background translates perfectly to venture dominance. [00:50:11]
Brett (Berson): First Round Capital partner referenced by Jack Altman during a discussion about the enduring value of "systems of record." [00:38:47]
Steve Jobs: Referenced for his famous Stanford commencement framework of "connecting the dots looking back." [00:49:06]
Models & Concepts
Qwen / Mistral: Advanced open-source AI models; Volpi points out their latest iterations are closed-source because the $50B+ training costs are too immense to give away. [00:24:54]
Llama / DeepSeek: Models used to illustrate the unsustainable economics of "giving away" frontier AI out of the kindness of an organization's heart. [00:26:30]
TPUs (Tensor Processing Units): Google's proprietary custom silicon, granting them a total exemption from the "Nvidia Tax" bottleneck. [00:35:09]
ASICs (Application-Specific Integrated Circuits): The future of AI hardware compute, where flexibility is sacrificed to hardcode neural weights directly into the silicon. [00:35:49]
8. The Bottomline (by AI)
The era of generic SaaS metrics, broad-stage venture capital, and Nvidia's unchecked general-compute monopoly is functionally over. The future belongs to highly specialized, capital-dense foundational incumbents at the bottom of the stack, and deeply embedded, service-heavy application companies (powered by Forward Deployed Engineers) at the top. For operators and investors alike, survival dictates abandoning the operational dogmas of the 2010s tech boom; you must either pivot legacy software to an AI-native architecture or face margin-crushing obsolescence. Watch the semiconductor space aggressively—the impending fracture from general GPUs to hard-coded, mission-specific ASICs like Cerebras and Etched will redraw the entire economic map of artificial intelligence.
Jun 13, 2026
This Week in Review | IPOs, US Inflation, ECB Rate Hike (June 12, 2026) | Fisher Investments
The June 12, 2026 edition of "This Week in Review" by Fisher Investments provides a high density macroeconomic analysis of three pivotal global market events: the landmark initial public offering IPO of SpaceX, the accelerating May U.S. co…
600 people
The number of employees at large, legacy VC firms hired just to "do spreadsheets"