The core thesis is that AI is still early, usage and model capability are compounding exponentially, and the market is underestimating how much that can keep pulling revenue, capex, and earnings higher. The near-term setup is volatile: AI stocks are tied to Fed moves, cloud growth, and megacap earnings on October 28, with positioning already bearish and a flush or squeeze possible. Beneath the swings, the case is that the market can stay cheap versus earnings while owning the infrastructure layers around the AI buildout. The sponsor pitch is VNRX, a high-risk epigenetics/diagnostics name with early cancer and sepsis data, animal tests, falling opex, and no signed revenue license yet.
Google token usage rose from about 10 trillion tokens a month in spring 2024 to more than 3.2 quadrillion by spring 2026 — over 300x in 2 years.
AI capability is described as compounding: since 2023, task-handling ability doubled about every 4 months, then about every 7 months in 2025; OpenAI’s best model in May 2024 handled 6-7 minute tasks, while Claude Methos preview by March 2026 handled tasks taking at least 16 hours.
Epic AI says the price of a given level of AI performance has fallen about 13x every year, and every 10x cost drop creates “a hundred new ways to use it,” lifting usage instead of just saving money.
Anthropic did $4.6 billion of revenue in all of 2025, $4.7 billion in Q1 2026, then 11.6 billion and 16.8 billion as presented; Bloomberg said OpenAI was running at more than $40 billion a year in August.
Microsoft, Alphabet, Amazon, and Meta are on track to spend about $730 billion on AI infrastructure this year; Wall Street had expected big tech to spend about $465 billion in 2026 in fall 2025, then 527 billion by November, and Goldman Sachs is cited at 1.2 trillion for 2027 and 1.4 trillion for 2028.
S&P 500 Q3 earnings are up 29.5% and tech earnings about 65%, while the S&P trades at about 18.7x next year’s earnings versus a 5-year average of 19.8x — profits are running ahead of prices, not the other way around.
The NASDAQ 100 is down 23% into April 2025, then down 12% into March 2026, then down 11% into July, then down about 7% since its record close on October 6.
AI stocks are said to be priced on growth rate, so a 10-point slowdown can wipe out almost 30% of where a company ends up; Coreweave is down about 38% since October 5 on a financing scare, not a demand scare.
Supply is described as lumpy because it needs chips, memory, power, permits, politics, and buildings with electricity, so revenue arrives in bursts when new capacity comes online; Wall Street is said to dislike and misunderstand that lumpy pattern.
Jensen Hong said on October 15 that this is “the early innings of the buildout of this new computing architecture” and warned it is “a very big mistake” to think only a couple of companies are funding the infrastructure.
A shock to oil, a hot inflation print, or a Fed hike tends to trigger selling in the most loved AI names; on September 16 the Fed hiked and the 10-year was said to be near 5%, hitting the biggest borrowers hardest.
The worst bearish calls are traced from 2023’s AI bubble call to 2024’s data wall argument to January 27, 2025’s DeepSeek-cost story that supposedly ended AI demand; since that close, NVDA is said to be up 78% after a CBS clip noted its nearly $600 billion one-day market-value loss and 17% drop.
Wall Street expects Q3 earnings growth of almost 30%, the market is described as cheaper than its own average while profits boom, and fear is said to spread because warning about downside gets more clicks and looks more astute than forecasting upside.
Gallup in July says 79% of Americans expect AI to reduce US jobs over the next 10 years; Reuters in June says 53% worry AI could cost them or someone in their household a job; Gallup in March says about 7 in 10 oppose an AI data center in their area.
Data center watch says at least 120 projects worth almost $200 billion were blocked or delayed by local opposition in the first half of 2026, while Census Bureau data says only 19.8% of American businesses used AI as of May 2026 and 37% of companies with 250+ employees did.
October 28 is the key near-term catalyst: at 2 PM Fed futures price an 85% chance of a hold and a 73% chance of another hike on December 9; at 2:30 Kevin Wars takes questions, then after the bell MSFT and GOOGL report, with 45% growth and $514 billion cloud backlog singled out as the numbers to watch.
On November 3, midterm season is framed as historically bullish for stocks: the S&P 500 has been higher 12 months after every midterm since 1950, up 19 for 19 on that measure and about 15% on average, with the average midterm-year drawdown about 17.5% from high to low.
The strategy is to own the layers the exponential curve runs through — chips, networking, power and cooling, data centers, and cyber security — while buying in stages, keeping dry powder, avoiding leverage, and holding through volatility.
VNRX is pitched as a small-cap epigenetics company building low-cost blood tests around nucleosomes to detect sepsis and cancer in people and cancer in dogs and cats, with headquarters in Henderson, Nevada and R&D in Belgium.
VNRX’s sponsored pipeline includes nu.qnets for sepsis, nu.q cancer, capture sec for human cancer, nu.qvet for dogs and cats, and nu.cover research kits; the sepsis pitch cites about 166 million cases a year and 31.5% of all deaths worldwide in 2021.
The sponsored cancer data says capture sec early results showed 49 of 49 cancers detected, including 23 stage 1 or 2, with no false positives among 21 healthy people; the results still need much larger confirmation studies.
The sponsored animal-data pitch says the K9 test is sold in more than 20 countries, a CAT prototype detected 86% of feline lymphas with 97% specificity, publication could unlock a $5 million milestone, and a CAT product could roughly double the animal market.
VNRX says it is in licensing discussions with more than a dozen diagnostics and liquid-biopsy companies, but no revenue-generating license has been signed; opex fell 32% in Q2, first-half 2026 revenue was 1.4 4 million up 112%, and it prepaid the remaining balance on a convertible note in September to reduce potential dilution.