Sep 25, 2025, 10:26 AM ET

The Impact of H1B Visas on Startups in the US & NVIDIA Invests $100BN Into OpenAI

NVIDIA is investing $100 billion in OpenAI, creating a potential feedback loop where OpenAI commits $300 billion to Oracle to purchase more NVIDIA chips.
OpenAI currently operates with only six major customers (including Microsoft, Google, Meta, and Oracle), yet these six entities are collectively spending "into oblivion" to dominate the AI space.
NVIDIA's market capitalization of approximately $4.5 trillion is supported by revenue concentration where six customers account for roughly 83% of total revenue.
The transaction allows Sam Altman to test whether AI scaling laws can sustain themselves with "three orders of magnitude" more compute than currently available, without an external "timeout" on funding.
Unlike previous tech booms (e.g., the dot-com era), current participants have sufficient balance sheet cash flow ($60 billion in projected fiscal 2025 free cash flow for NVIDIA) to avoid immediate solvency collapse.
NVIDIA has authorized a $60 billion stock buyback program, a figure equivalent to its projected annual free cash flow, which critics argue is aggressive financial engineering during a bull market.
Data labeling companies (e.g., Scale AI, Appen) show extreme revenue concentration, with two customers comprising 55% of revenue for major providers in the sector.

In 2025, 75% of all venture capital dollars flowed into just 19 companies, indicating a hyper-concentration of capital similar to the "MAG7" trend in public markets.
Venture firms are increasingly bypassing rigorous due diligence on hot AI deals, often issuing term sheets on Saturdays with the mindset that "the only thing you can lose is one extra million."
"Founder-friendly" has shifted from a differentiator to a baseline requirement; true founder-friendliness is now defined by providing capital when no other investors are present or actively recruiting executives, rather than just offering moral support.
Investors report a return to pre-2021 skepticism, with less willingness to bet on stories that cannot substantiate margins, churn, or retention metrics at scale.
Early-stage success is highly correlated with future VC success, driven by both referral networks and the increased confidence ("stomach to roll the dice") gained from previous wins.
The "Triple, Triple, Double, Double" (3x, 3x, 2x) growth benchmark is becoming insufficient for securing funding; investors now require companies to demonstrate growth rates significantly exceeding this threshold to justify valuations.

Public market valuations are heavily dependent on consumer adoption and spending decisions, with ChatGPT estimated to serve 10% of the world's adult population as a weekly active user.
The market is currently pricing assets based on forward-looking potential rather than immediate fundamentals, with a consensus that 2021-era valuations are "flushed down the toilet" as a relevant metric.
Investors are debating whether to exit positions in winners like OpenAI or Anthropic; the consensus leans toward holding to defer capital gains taxes while the potential for a 2x return on paper increases.
The "risk aversion" of investors shifts with net worth; founders and investors with significant personal wealth are more likely to "let their winners ride" compared to emerging managers who must realize DPI (Distributions to Paid-In Capital).
A "froth" indicator has emerged where Limited Partners (LPs) are publicly bragging about returns on LinkedIn, a signal historically associated with the peak of the 2021 bubble.

Navan is proceeding with an IPO despite lacking profitability to avoid becoming the "third player" in a crowded travel/expense management sector after competitors Brex and Ramp go public.
Navan's strategy involves rebranding from a vertical travel tool to a horizontal platform to justify a higher multiple, though analysts note the core revenue remains tied to travel booking rather than pure software.
Atlassian's recent acquisitions (e.g., Dataloom, Exa) are viewed as defensive measures to maintain relevance and defend market cap rather than a transformative shift to AI leadership.
The "lockup" period for IPO shares is typically six months, but early liquidity can sometimes be achieved via secondary sales if the stock trades above the IPO price, or via distributing shares to LPs.
Some investors, like Sequoia partners Mark Stevens and Tench Cox, have held their NVIDIA shares since the 1997 IPO without selling, illustrating the long-term value of undivested early stakes in dominant platforms.

The proposed $100,000 fee for new H-1B visas is expected to have a modest negative impact on early-stage startups, though companies will likely pivot to O-1 visa categories for founders.
Critics argue the fee is a crude proxy for skill selection, whereas a rational policy would attach visas directly to STEM degrees or specific high-impact projects.
There is skepticism that the H-1B increase will significantly alter the ecosystem, as major tech firms can absorb the costs and startups can find alternative talent acquisition strategies.
The "Kalshi Quickfire" segment noted a 60-day outlook for a potential US-China TikTok deal, driven by the need to resolve tariff posturing and trade tensions.

Smart glasses are predicted to fail as a mass-market consumer product ("zero percent chance") because they lack a compelling "real world" problem to solve compared to the smartphone paradigm.
Conversely, there is hope for a new computing paradigm via devices like those rumored from Jony Ive, which could unify computer, vision, and living spaces, though this remains a risky, unproven bet.
Notion is projected to reach $500M in ARR (Annual Recurring Revenue), marking a re-acceleration for a mid-tier SaaS company, though it does not meet the "doubling" growth rates of the early 2020s.
Investors are advised to focus on predicting the next 18-24 month financing round rather than attempting to forecast valuation multiples five years out.
The prevailing view is that the AI CapEx boom will continue until a clear "wall" is hit where marginal capital no longer yields proportional returns in intelligence or efficiency.