Anthropic Inference Costs Skyrocket |TikTok Deal Closes |The IPO Market:Wealthfront & EquipmentShare

Brex Acquisition by Capital One

• Capital One acquired Brex for $5.15 billion in a deal structure of 50% cash and 50% shares.
• Founders and early employees achieved a "heroic" financial outcome, though late-stage investors face dilution relative to the company's $12 billion valuation in 2021.
• The deal reflects the "hubristic financing" phenomenon where companies raised at peak valuations to secure growth momentum, accepting the risk of lower exit multiples later.
• Capital One gains a structural competitive advantage by integrating Brex with its recently acquired Discover Card network, allowing it to monetize interchange fees internally rather than relying on third-party Visa/Mastercard rails.
• The transaction sets a market benchmark of approximately 7x revenue for a mature fintech, creating pressure on competitors like Ramp to justify higher multiples (currently ~30x) despite faster growth rates.
• Founders of Ramp face a difficult choice: grind for a public offering where valuation may converge with Brex's normalized multiple, or risk remaining private to preserve a higher theoretical valuation that the public market may not support.

TikTok Divestiture to US Investors

• A deal finalized allowing US investors to own 80% of TikTok while the algorithm remains under Chinese ownership.
• The acquisition price is approximately 1x to 1x revenue, described as "wildly cheap" given TikTok's $15–16 billion US revenue run rate.
• Major venture firms (Andreessen, Sequoia, Lightspeed) notably abstained from the deal, prompting concerns about hidden risks or structural challenges despite the apparent arbitrage.
• The deal echoes the successful Andreessen/Horowitz acquisition of Skype from eBay, where a distressed asset was cleaned up and sold to Microsoft for a 3x return.
• Sovereign wealth funds (e.g., UAE) and Oracle participated, signaling strong institutional appetite for assets with guaranteed demand in the US market.

Anthropic Inference Costs and AI Economics

• Anthropic's inference costs rose 23% higher than expected, yet the company improved gross margins from -94% the previous year to +40% currently.
• Demand for inference tokens is accelerating rather than deflating; as models improve, consumption increases, potentially reaching "infinite" demand in a 24/7 knowledge economy.
• The AI cost structure is shifting from traditional infrastructure (AWS) to inference costs, which can constitute 50–70% of revenue for coding and agent-based applications.
• Mid-tier SaaS companies face a critical capital trap: they must raise funds to match the token consumption of better-capitalized competitors (like Open Evidence or Harvey) to remain competitive.
• The viable path for legacy SaaS to survive AI disruption is to deliver agent value with an ROI so high (e.g., replacing 20 employees) that they can charge premium fees covering the massive inference costs.
• Semiconductor manufacturers (TSMC, NVIDIA) and hyperscalers are increasing CapEx budgets significantly, signaling confidence that compute demand remains real for the next 12–24 months despite the widening gap between CapEx and current app revenues.
• While the AI bubble is expected to eventually burst due to overinvestment, experts predict no immediate downturn in the next two years given the "bottom-up" conviction of hardware makers.

Open Evidence Valuation and Market Size

• Open Evidence raised $12 billion at a 12x valuation step-up, backed by Thrive Capital and DST.
• The company holds a commanding position in the doctor decision-support niche, leveraging HIPAA compliance and exclusive relationships with medical journals to replace traditional pharmaceutical reps.
• Market skepticism centers on the Total Addressable Market (TAM): direct-to-doctor advertising is a ~$3 billion market, implying Open Evidence must expand services beyond pharma ads or steal budget from in-person sales to justify a $5B+ revenue run rate.
• The deal is viewed as a "hubristic" entry point, similar to Brex's 2021 raise, where early investors may eventually face a "1x" return if the company fails to scale revenue to match the hype.
• Andreessen Horowitz reported that two-thirds of private AI revenue comes from their portfolio, with OpenAI accounting for 40–50% of that total, raising questions about whether the AI sector constitutes a viable "asset class" justifying 10x more capital deployment.

Public Market Trends: IPOs and SaaS Dynamics

• Equipment Share's successful IPO (pop 33%, $8B market cap) signals that profitable, scaled companies with 40%+ growth can access public markets, whereas sub-scale or loss-making companies (e.g., Wealthfront) face immediate devaluation.
• Wealthfront's IPO performance (trading down ~36% from issuance) highlights a "perilous zone" between $1–2B market caps where liquidity is poor and the company is "barely public."
• Ethos Health's IPO at a $1.3B valuation (down from a $2.7B private valuation) represents a capitulation by investors seeking liquidity after failing to secure a strategic M&A exit.
• A Salesforce Army contract awarded a $5.6B, 10-year deal reinforces that legacy SaaS systems of record are not being replaced by AI agents but are evolving to integrate AI-driven sales forces.
• SaaS growth faces structural headwinds including seat contraction, aggressive historical price increases (40% over 3 years) crowding out upsell potential, and workforce reductions.
• The consensus is that SaaS is not dead but is entering a period of "boring" stability where growth will be slower, and the old model of 100%+ net revenue retention is unlikely to return.