🧭 The Venture Paradox: Can a Firm Be Forgiven When the Partners Change?
A fresh round of soul-searching is rippling through venture after renewed debate over whether it’s time to “forgive” Benchmark for the ousting of Travis Kalanick. The question isn’t simply reputational. It goes to the heart of how partnerships evolve and whether brand legacies can be re-written as the roster turns over — the classic Ship of Theseus paradox applied to VC.
- On June 20, 2017, Kalanick resigned as Uber CEO under investor pressure that included Benchmark.
- At that time, Benchmark’s equal GP roster included Bill Gurley, Eric Visra, Matt Kohler, Mitch Laskkey, Peter Fenton, and Sarah Tavl. Today, only Peter and Eric remain, with Chattton, Ev Randall, and Jack Altman joining post-Uber. In other words, about one-third of the 2017 partnership remains.
“How many partners need to be at the firm until we can call this a ship of Thes(i)us?”
The unsettled point: if — or when — the last two legacy partners depart, does the stigma depart with them? Or is the brand the “ship” that persists, boards replaced or not?
🚗 Autonomy and the Uber What-Ifs
Uber’s trajectory under different leadership remains a live counterfactual. The argument centers on self-driving and whether Uber forfeited a defining second act by shelving its in-house push.
- Uber today is cited at ~$150 billion. Lyft has fallen to ~$5 billion.
- Waymo was cited as having been valued in February at $126 billion.
“It’s hard to imagine Uber being worth less than something like a trillion dollars today [if Travis had stayed].”
“In my opinion, Gurley single-handedly destroyed hundreds of billions in value... a Teslasized win 500 billion plus for everyone including Benchmark’s LPs.”
Current Uber strategy leans on partnerships with AV players rather than full-stack internal IP — a calculated trade that has contained risk and capital intensity but, in some eyes, capped the upside and narrative premium that a “one-of-one” founder plus proprietary autonomy might have commanded.
🧪 Incentives, Power Laws, and a Steelman for Benchmark
There’s an uncomfortable, if plausible, incentive-based read on the 2017 call: the power law pressure of an equal partnership sitting on one mega-outcome.
“Every partner at Benchmark... was going to make a clean $1 billion.”
Layer on boycotts, media pile-ons, and the optics risk of a zero in a highly public name — and it’s easy to imagine a flight-to-liquidity instinct overriding principle. Not moral clarity; just raw exposure. As that narrative goes, fear drove the decision more than any single “egregious” act. The reputational hit since has likely cost access to founders and deals, even if returns have remained strong.
🎬 Sora’s Sunset and the Economics of AI Video
OpenAI’s Sora app is being phased out, with functionality expected to consolidate into a single interface. The move underscores a hard truth: consumer-scale AI video is compute-heavy, expensive, and tough to retain without seamless, instant loops.
- Millions of people created content on Sora, but consolidation signals discipline and focus.
- Google’s video tooling offers a telling benchmark: an entry tier at $250/month (moving to $500), with brutal rate limits — as few as three generations per day in some experiences.
“Rate limits kill retention like nothing… imagine if TikTok was like after five minutes you have to close the app and come back in 20 minutes.”
Two deeper takeaways:
- Tool-to-network is a high wall. File outputs (MP4, etc.) are uploadable on day one to TikTok/Instagram/YouTube, erasing greenfield advantages for new networks. The classic “come for the tool, stay for the network” playbook struggles when incumbents absorb the format instantly.
- Compute flows to where tokens are most valuable. Enterprise codegen, knowledge retrieval, and workflow automation outcompete open-ended video generation on ROI. Expect labs to keep shifting cycles to higher-value tokens.
“Probably 5% of the text that I read is LLM.”
Consumer AI video adoption looks like a long diffusion curve: from one-off viral novelties to steady niches. Early traction is showing up in formats like a TikTok-native, AI-generated reality series and even a fully AI-generated podcast topping charts — evidence of product-market fit in pockets, not a universal doom-loop.
🎥 Marketplaces in the Middle: Fiverr’s ‘AI Directors’ Bet
Fiverr is leaning into an “AI directors” positioning — promoting creators who orchestrate multi-model pipelines, fine-tunes, and tooling stacks to deliver brand assets at a fraction of traditional production cost. A provocative billboard even triggered a fender-bender, per local TV reportage.
Strategically, the bet acknowledges two realities:
- Orchestration value is real: stitching models, fine-tunes, and video pipelines is non-trivial and valuable to brands.
- Disintermediation risk persists: star freelancers can be contacted directly, a long-standing headwind for labor marketplaces that differs from ride-hailing-style liquidity.
🛫 Airlines vs. Bizjets: A New Middle Seat Strategy
Airline innovation is nibbling at private-aviation use cases on long-hauls. United introduced the United Relax row — three adjacent economy seats with adjustable leg rests that create a lie-flat surface, plus a mattress pad, blanket, and two pillows (with a plushy for kids). The product is slated to roll out starting next year on more than 200 Boeing 787s and 777s. A similar concept launched at Air New Zealand in 2010.
Meanwhile, Bombardier shares are reported to be down 10% over the past month — a reminder that premium economy innovations can pressure marginal private travel demand at the edges.
🚀 SpaceX S‑1 Watch: The AI Unit Economics Angle
Rumors swirl that SpaceX may file for an IPO as soon as this week, stoking anticipation around disclosures that could surface in an S‑1 — especially for xAI and Grok’s inference economics. One focal point: whether consumer inference is being served profitably at observed price points.
- There’s a stark market spread between AI service pricing tiers — for instance, a “wild difference” between $15 per million tokens and $2 per million tokens — which will shape margin narratives for any AI adjacency.
“I would put the April 20th IPO at maybe like, you know, 30% and then the ticker maybe down at 15%.”
Ticker speculation aside, any granular breakout of AI costs, utilization, and monetization would be a first look at the true unit economics of consumer-scale inference inside an at-scale operator.
Key Quotes
- “It’s hard to imagine Uber being worth less than something like a trillion dollars today [if Travis had stayed].”
- “Waymo was valued in February of this year at 126 billion.”
- “Every partner at Benchmark... was going to make a clean $1 billion.”
- “Probably 5% of the text that I read is LLM.”
- “Three a day.” (on rate limits for AI video generation)
- “There’s a wild difference between charging $15 per million tokens versus $2 per million tokens.”
- “United Relax row... three adjacent United Economy seats... mattress pad, blanket, and two pillows... available starting next year on more than 200 of the 787s and 777s.”
What to Watch
- Benchmark’s next chapter: Partner turnover vs. brand persistence; whether founder sentiment warms as the Ship of Theseus nears completion.
- Uber’s AV posture: Partnership model vs. proprietary bets; whether autonomy economics become a valuation overhang or catalyst.
- AI video reallocation: Fewer consumer-facing skews, more enterprise-grade pipelines; continued rate-limit and pricing signals from hyperscalers.
- Creator marketplaces: Whether orchestration-as-a-service can overcome disintermediation risk.
- Aviation substitution: Uptake of lie-flat economy innovations and any further pressure on bizjet OEMs.
- SpaceX filings: Any visibility into xAI revenues, costs, and inference margins; real pricing/margin signals across tokens.
