Jensen Huang just set the AI infrastructure agenda for the next two years

The Big Story: Nvidia's GTC Sets the AI Compute Agenda

Jensen Huang took the stage at San Jose's SAP Center yesterday before 30,000 attendees from 190 countries, and the headline was the full technical unveiling of the Vera Rubin architecture — the successor to Blackwell, built on TSMC's 3nm process with HBM4 memory and designed specifically for the agentic AI and inference workloads that are now where the real money is.

A few numbers worth understanding: Rubin reportedly cuts the token cost for advanced reasoning and large mixture-of-experts model inference to roughly one-tenth of what Blackwell required, while needing only one-quarter of the GPUs for equivalent training workloads. That's a meaningful cost reduction for the hyperscalers — and potentially for the private companies whose valuations are partly a bet on AI compute economics improving.

Huang also put Feynman officially on the roadmap for 2028, with early details pointing to TSMC's A16 1.6nm process and silicon photonics for data transmission. And on the software side, Nvidia is reportedly launching NemoClaw — an open-source platform for deploying enterprise AI agents — which would extend Nvidia's ambitions from chips into the software layer companies use to run those chips.

What this means for private market investors: The companies on Augment's Power 20 — Anthropic, OpenAI, xAI, Databricks — are all downstream beneficiaries of cheaper, faster compute. Rubin's efficiency improvements suggest the cost curves that matter for AI business models are continuing to move in the right direction, even as the capital expenditure required to stay current keeps climbing.

Nvidia GTC live coverage

Meta's "Year of Efficiency 2.0": 20% Workforce Cuts Reported

Reuters reported Friday that Meta is planning layoffs that may affect 20% or more of its workforce — potentially more than 15,000 employees, based on the company's headcount of roughly 79,000 at year-end 2025. No date has been set and the scope isn't finalized; a Meta spokesperson called the report "speculative reporting about theoretical approaches."

The stated rationale is familiar by now: offsetting AI infrastructure costs while preparing for efficiency gains from AI-assisted workers. Meta has guided to capital expenditure of $115–135 billion in 2026 — roughly double last year's spend — to build out the data center infrastructure for its AI ambitions. The workforce reduction, if it happens at scale, is the cost side of that equation.

Jefferies analysts noted the implication: "If Meta is willing to reduce headcount at this scale while ramping AI investment, we think it signals a broader shift — AI is increasingly driving productivity. That has important implications not just for Meta, but across the broader internet/software landscape."

For context, Amazon confirmed 16,000 job cuts in January, Block cut roughly 40% of its staff in February explicitly citing AI productivity gains, and Oracle has reportedly been contemplating cuts of 20,000–30,000 to fund data center expansion. The pattern is becoming consistent enough to call a trend.

Reuters report on Meta layoffs

Quick Takes

The U.S. Army awards Anduril a 10-year, up-to-$20B contract for software, hardware, and services — The deal includes a five-year optional ordering period and is one of the largest defense tech contracts awarded to a private company. Anduril, last valued at $28B, remains one of the most active names in secondary markets.

Travis Kalanick comes out of stealth with Atoms, an industrial robotics venture — The Uber co-founder published a 1,700-word manifesto this week detailing an eight-year effort building industrial robotics. Worth reading if you want to understand where the CloudKitchens years were actually heading.

Legora raises $550M Series D at a valuation not disclosed, led by Accel — The Stockholm-based AI legal platform raised one of the larger rounds of the week. Legal tech is emerging as a category with real enterprise adoption, not just demos.

Eridu raises $200M Series A — networking platform aimed at enterprise AI infrastructure — A $200M Series A is unusually large; this is one to watch in the emerging AI infrastructure layer.

Paul

Preferred vs. Common Stock

Private companies typically issue preferred stock to institutional investors and common stock to employees. Preferred shares carry rights common shares don't — liquidation preference, anti-dilution protection, sometimes a dividend. In a secondary transaction, buyers are usually acquiring common shares, which means they're lower in the capital structure than institutional investors. Understanding the difference matters when evaluating price and risk.

What We’re Watching

1. GTC announcements through Thursday — Nvidia's event runs through March 19 with additional sessions on physical AI, agentic systems, and robotics. Major partnership announcements are expected to continue through the week.
2. Meta's actual layoff announcement — The Reuters report was attributed to unnamed sources; Meta has not confirmed scale or timing. If the 20% figure is accurate, the market reaction will matter as a signal for how investors are pricing the AI cost vs. efficiency trade-off.
3. Anthropic lawsuit progress — The company is seeking a preliminary injunction to block the supply chain designation. Given that the DoD is simultaneously using Claude in active operations, the legal posture is unusual. How courts handle the "mission critical" carve-out will have implications beyond this case.

Let’s have a great week out there.

~

Paul

See what's trading on the secondary market: augment.market

Augment Markets, Inc. is a technology company offering software and data services with securities-related services offered through its wholly-owned but separately managed subsidiary Augment Capital, LLC, Member of FINRA/ SIPC.

Important Disclosures: This material has been prepared for informational purposes only. None of the information provided represents an offer or the solicitation of an offer to buy or sell any security. The information provided does not constitute investment, legal, tax, or accounting advice. You should consult with qualified professionals before making any investment decisions. Investing in private securities involves substantial risk, including the potential loss of principal. Private securities are typically illiquid, have limited pricing transparency, and often require longer holding periods. These investments are available exclusively to qualified accredited investors and offer no guarantee of returns. Additionally, past performance of private securities does not indicate or predict future results.

‍