DeepSeek Rattles Markets + SailPoint's Going Public… again? | E2077

Episode Summary

Executive Summary: The episode centers on DeepSeek’s low-cost, high-quality AI model and its shockwaves through markets, especially NVIDIA’s selloff. The hosts debate whether cheaper AI reduces hardware demand or expands it via Jevons’ paradox, while also discussing open source, censorship, security/privacy, startup shutdowns, M&A, and how AI could reshape workflows, browsers, and company size.

Main Topics: DeepSeek shocks AI markets and NVIDIA (Priority: 5/5): The hosts discuss how DeepSeek’s model release challenged assumptions about AI progress requiring massive capex, triggering a major selloff in NVIDIA and broader AI stocks. Jevons’ paradox and AI demand expansion (Priority: 5/5): They argue that cheaper AI may not reduce compute demand; instead, like other technologies, it could increase usage and create even more applications and infrastructure needs. Open source, model replication, and geopolitics (Priority: 4/5): The conversation explores how open-source releases accelerate adoption, make censorship easier to bypass locally, and reduce the effectiveness of chip export restrictions against China. Startup shutdowns, burn, and M&A (Priority: 4/5): The hosts review rising startup shutdown data, attributing many failures to companies that couldn’t adjust burn after the zero-rate era and hoping for more acquisitions and acqui-hires. AI agents, browsers, and workflow automation (Priority: 4/5): They predict AI will increasingly act inside browsers and across repetitive tasks, reshaping productivity software, operating systems, and even content production workflows. Private equity and public market structure (Priority: 3/5): A side discussion on SailPoint highlights how private equity can load companies with debt before relisting them, raising questions about long-term value creation versus extraction.

Key Arguments: DeepSeek suggests frontier-quality AI may be built with far less compute and money than previously assumed, undermining the narrative that only enormous capex can win. If AI becomes cheaper, demand may rise rather than fall; more startups and enterprises will find uses for it, echoing Jevons’ paradox. Open-source AI will diffuse globally, so useful techniques from DeepSeek are likely to be absorbed by U.S. labs and startups regardless of geopolitics. Export controls and hosted-model restrictions are only partial barriers because local deployment can bypass censorship and platform limitations. The current AI market may push closed-source products to justify their valuations by demonstrating real utility and demand. Many startup shutdowns are the result of companies failing to right-size burn after peak-ZERP funding, hoping for rescue rounds that never arrived. Future AI value may accrue less to model providers alone and more to workflows, browsers, enterprise tools, and agents that sit on top of models. Private equity can improve companies operationally, but it can also saddle them with heavy debt and extract value before relisting them.

Data Points: DeepSeek training cost: $6 million - Referenced as the reported cost in the white paper; the hosts note debate over what was included. Project Stargate planned spend: $100 billion now, up to $500 billion - Used as an example of the massive capex assumptions DeepSeek is challenging. Meta 2025 capex: $60 billion to $65 billion - Cited as another huge data-center investment now looking less certain. NASDAQ move: Down 5% - Jason cites the market reaction after the DeepSeek news. NVIDIA market cap drop: About $150 billion deleted in a day - The hosts estimate the scale of the selloff tied to the AI trade reversal. NVIDIA share price move: Roughly $150 to $120; nearly 17% down - Used to illustrate the severity of the stock decline during the episode. Carta shutdown increase: 25.6% increase in 2024 vs. 2023 - Referenced in the startup shutdown discussion. AngelList shutdown count: 364 shutdowns, up 56% - Another data point showing a rise in failures. SailPoint ARR: $813 million as of October 2024 - Used to frame the company’s current scale ahead of its return to public markets. SailPoint interest expense: $140 million in three quarters - Highlighted as a result of debt taken on during the take-private deal. SailPoint term loan: $1.59 billion - The debt used to fund the private-equity transaction. SailPoint interest rate: 14% - The hosts note the unusually high cost of the debt, like a corporate credit-card rate. LinkedIn hiring stats: 1 billion members; 72% of small businesses; 86% qualified match within 24 hours - Promotional sponsor read, but it appears in the transcript with specific claims.

Pivotal Quotes: "The AI trade giveth and the AI trade taketh away." — Jason: Summarizing the market reversal after DeepSeek’s announcement. "When technological progress increases the efficiency of something, people use more of it." — Alex: Explaining Jevons’ paradox as the core frame for why cheaper AI may increase demand. "The biggest point that makes is a little bit less about China versus US, but about the fact that we are nowhere near having wrung out all the possible gains." — Jason: Arguing that DeepSeek reveals how much optimization headroom still exists in AI.

Implications: Listeners should expect cheaper AI to accelerate adoption, not kill the market. The winners may shift from pure model makers to tooling, browsers, security, and workflow automation, while investors rethink capex, valuations, and startup survival.