Instinct's $10B Talks: The VC Fund Math Behind It

Instinct is reportedly in talks to raise up to $1 billion at a valuation of up to $10 billion, according to The Information. The round hasn't closed. But the exercise of modeling it out is more interesting than the headline number itself: it shows how fast one early check can come to look like it's worth more than the entire fund it came from — on paper, before a single dollar comes back to an LP.

We covered the raise talks on Pulse when they broke, and flagged the cap table math as the thing actually worth checking before the valuation number. This is that math: a first-principles dilution model of what Conviction's and Greenoaks' early checks could be worth if the reported round closes at $10 billion, built from the ground up rather than reused from an earlier estimate, with every assumption shown.

Instinct's Funding Trajectory, From $50M to a Reported $10B

Instinct, the personal AI assistant built by Spear Street Technology and founded by Noah Shinn, has moved through five distinct valuation marks in under a year. Forbes reported that Pranav Reddy of Conviction led an early round valuing the company at $50 million, with Neil Mehta of Greenoaks also an early backer. By April 2026, Forge's data on the round shows roughly $17.3 million raised across two seed tranches at an approximately $147.74 million post-money valuation. In early August, Mamoon Hamid of Kleiner Perkins led a $75 million Series A that Forge marks at roughly $567.76 million post-money (Forbes and other outlets describe this more loosely as “over $500 million”). On August 26, TechCrunch reported a $250 million Series B co-led by Benchmark and Index Ventures at a $2.5 billion valuation, bringing total funding to $350 million. Some reporting on the prior round has used $2.25 billion instead of $2.5 billion for that same close — we use $2.5 billion throughout, the figure TechCrunch, Forbes, and Yahoo Finance converge on.

Two and a half weeks later, Instinct was reportedly back at the table. The Information first reported on September 10 that Instinct was in talks to raise up to $1 billion as compute costs strained its ability to serve a user base that had crossed 100,000 — the product reportedly sometimes tells users it's operating at full capacity and that “responses may be slower.” By mid-September, that had firmed into talks for a valuation of roughly $10 billion, with Sequoia Capital and Benchmark reportedly in discussions to lead; some reports have also named Coatue as a possible participant, though that detail is less independently corroborated than the Sequoia and Benchmark involvement. None of this has closed.

What's New Here: Two Early Checks, Not One

Most public reporting, including our own coverage of the Series B, treats Conviction and Greenoaks as co-investors in a single $50 million seed round. Forbes' reporting supports Conviction leading at a $50 million valuation and Greenoaks being an early backer, but doesn't specify whether Greenoaks invested at that same mark or a later one.

For this piece, more granular detail was provided directly: Conviction invested $5 million at a $50 million valuation, and Greenoaks followed shortly afterward with $10 million at a $100 million valuation — a distinct, higher mark, not the same round. That distinction matters for the math below, because it means Conviction absorbed one more round of dilution than Greenoaks did before the two funds otherwise moved through the same subsequent rounds together. We're flagging this explicitly: treat the $5M/$50M and $10M/$100M figures as information provided for this analysis, not as independently confirmed public facts, unless noted otherwise.

Building the Cap Table: A Simple Dilution Model

Instinct's actual cap table isn't public, so every ownership and paper-value figure below is an illustrative model, not a disclosed fact. The assumptions, stated plainly:

• Every “valuation at investment” figure is treated as post-money — the standard convention when reporting says a round “valued the company at $X.”
• Each subsequent financing dilutes existing holders by a flat 20%, with no pro rata participation by the earlier investor and no secondary sales.
• No option-pool changes, no additional securities, and liquidation preferences are ignored for this simple common-equivalent mark.
• A hypothetical $10 billion close is modeled purely to illustrate the mechanics — it is not a claim that Instinct is worth $10 billion today.

The financing sequence, and who each round dilutes: Conviction's entry ($50M) is diluted by everything after it — Greenoaks' entry ($100M), the April seed close ($147.74M), the Series A ($567.76M), the Series B ($2.5B), and the hypothetical $10B round. That's five dilution events. Greenoaks enters at the $100M mark itself, so it isn't diluted by its own round — it absorbs the remaining four: the April seed, Series A, Series B, and the hypothetical round. Kleiner Perkins enters at the Series A and is diluted by two subsequent rounds (Series B, hypothetical). Benchmark and Index enter at the Series B and are diluted by one (the hypothetical round only).

At 10% entry ownership each ($5M of $50M, $10M of $100M) and 0.8 raised to the power of the dilution-event count, Conviction models to 10% × 0.8⁵ = 3.28% ownership at the $10B mark; Greenoaks to 10% × 0.8⁴ = 4.10%. We are deliberately not reusing a widely-cited ~$512M estimate for Conviction's position — rebuilding the dilution count from scratch, event by event, is what produces the $327.68M figure below instead, and the difference is the point: get the dilution-event count wrong by even one round and the paper value swings by 25%.

What This Would Mean Against Conviction's Fund

Conviction, Sarah Guo's AI-focused firm, closed a $230 million second fund in early 2025, per TechCrunch and Forbes — more than double its $101 million debut fund. Conviction has since raised further vehicles, but the $230 million fund was the active vehicle at the time of an early-2026 investment, which makes it the relevant comparison here.

A single position modeled at $327.7 million is worth roughly 1.4x the size of that entire $230 million fund — on paper. That is not the same as saying the position “returned the fund.” Nothing has been distributed. It means one early check, if this mark holds and the position is eventually sold at anything close to it, would be worth more on paper than every dollar LPs put into the fund combined — a potential paper fund returner, not a realized one. (Greenoaks' current fund size wasn't independently sourced for this piece, so we aren't running the same comparison for its position.)

Same Company, Very Different Position Sizes

The contrast with the later-round investors is exactly why entry timing matters more than headline valuation multiples. Kleiner Perkins raised $3.5 billion across two new AI-focused funds in March 2026 — a $1 billion early-stage vehicle (its 22nd) and a $2.5 billion growth fund, its largest raise since the firm's 1972 founding, per TechCrunch and Bloomberg. Against a modeled $845.6 million position from the $1 billion early-stage fund, that's a large, concentrated bet (roughly 85% of the fund on paper) but not a multiple-of-fund outcome the way Conviction's smaller, earlier check is.

Benchmark is the clearest illustration of scale mattering more than multiple. Benchmark closed $2 billion across two funds in June 2026 — a $750 million flagship early-stage fund and its first-ever dedicated growth fund at $1.25 billion, a structural break from the roughly $425 million fund size it held for two decades, per TechCrunch and Axios. We don't know which of the two funds wrote Instinct's Series B check, or how the round split between Benchmark and Index Ventures. Even in the most generous illustrative case — the full $250 million from Benchmark alone, out of its $750 million early-stage fund — a modeled $800 million paper value is meaningful but nowhere near fund-transforming. The same company can be a fund-defining position for one investor and a solid, unremarkable one for another, purely as a function of when and how much they wrote the check.

MOIC, TVPI, DPI, and IRR — What Each Actually Measures

MOIC (Multiple on Invested Capital) is current value divided by capital invested, for a single position — the $65.5x figure above. TVPI (Total Value to Paid-In Capital) is the fund-level equivalent: realized distributions plus remaining unrealized value, over capital called from LPs. DPI (Distributions to Paid-In Capital) counts only cash or stock actually paid out. A position can carry a 65x MOIC and contribute a fund-level TVPI bump while its DPI contribution sits at exactly zero, which is the case modeled here. For the fuller picture of how that gap has widened across the industry, we've written separately about what current TVPI-vs-DPI benchmarks actually show.

IRR is where this kind of math gets genuinely misleading if you're not careful. Over exactly one year with a single initial investment and a single terminal value, IRR simplifies to MOIC minus one. Modeled over roughly the year since Conviction's entry, that's a MOIC of 65.5x producing an annualized IRR figure north of 6,400% — a number that is mathematically consistent but practically meaningless as a statement about repeatable annual returns. We're showing it once, here, specifically to illustrate why very short holding periods on very large markups should be treated as an artifact of annualization, not a real performance claim.

A Markup Is Not a Distribution

None of the paper values above are cash. Under fair-value accounting guidance, a fund marks a position to the price of its most recent priced round — so if Instinct's reported $10B talks close, Conviction's and Greenoaks' carrying values move immediately, before either fund has sold a single share. That's standard practice, not a red flag on its own; we've covered how that mark-up culture plays out across the industry more broadly. The distinction that matters is simpler than it sounds: TVPI can move on a headline; DPI only moves when an LP actually gets paid.

It would also be a mistake to assume a markup like this automatically makes LPs more willing to commit to a manager's next fund. Sophisticated LPs increasingly discount recent, unrealized marks precisely because they've seen this pattern before — a single hot position inflating TVPI without moving DPI at all. What a mark like this does change is the story a fund can tell while raising: a GP with a modeled fund-returning position, even unrealized, has a very different pitch than one without. Whether that story converts into actual fundraising success is a separate question from whether the mark itself is real money, and it isn't, yet.

Why Instinct Might Not Need to Be Profitable Yet

The secondary story here is about what this much capital access does to a young company's incentives. For most startups, profitability matters early because capital is scarce and every dollar of burn has to be justified. Instinct, if it successfully raises another $1 billion, is operating under a different constraint entirely: reported user growth past 100,000 and compute capacity that's reportedly already strained enough that the product tells users to expect slower responses. That combination — explosive demand and investors willing to fund it — changes the calculus.

In that specific situation, maximizing near-term profitability can conflict with maximizing growth. The more coherent strategy, while capital is this available, is closer to: get distribution, build something people actually use, figure out monetization later. Most startups don't have that luxury — they need a path to durable economics well before they have 100,000 engaged users, because no one is going to hand them $1 billion to figure it out along the way. Instinct, for now, is being underwritten on the bet that it eventually will need one anyway; a compute-driven raise is historically as often a warning sign as a growth signal, and what happens to unit economics once (or if) growth slows is the thing worth watching next, not the valuation number itself.