Buy Heavy Into the Dip? Why a 10-20% September Pullback Might Not End the AI Bull Run

The core pitch in one line is striking: the AI wave has delivered roughly nine-fold gains in Nvidia and about 45 times your money in SanDisk, and with the calendar turning to September a 10-20% pullback looks imminent — yet the host frames it as a heavy buying window, not a reason to run. The Nasdaq remains about 3% below its June peak and has chopped sideways since May, so the message is to plan the dip rather than fear it, with a playbook for why it happens and how to know when the cycle truly ends.

Why September matters is backed by history: in just over 30 years the market has closed higher less than half the Septembers, averaging about -0.84%, the only month whose average return is negative across all four major U.S. indexes. This year the setup looks heavier because stocks ripped 31% in two months off the April low to a fresh peak, S&P 500 companies posted 47% earnings growth in the second quarter — the strongest since the 2021 stimulus era — and valuation pushed to about 26 times earnings. That is not a historic peak compared with 28 times for much of the past two years, but it sits roughly 10% above the 10-year average and no longer looks cheap.

How likely is a shakeout? In 36 of the last 50 years the market has seen a 5-20% drawdown before eventually recovering, nearly three in four years. A 5-10% slip in the broad market often translates into 10-20% or more in the hottest names because high-beta leaders swing harder. The host treats a 3% dip as a near certainty and argues the real pain will cluster in crowded AI flyers that ran the hardest off the April low.

When does the bull actually turn into a bear? The answer is tied not to headlines but to signed financials: investors should watch the first audited quarterly prints from private leaders OpenAI and Anthropic. IDC's estimate of a $22.5 trillion AI-linked economic prize and the eye-popping 45x and nine-fold price moves keep enthusiasm high, yet durability will be judged on billable profit. Until those statements arrive, news flow is noise.

The warning underneath is about pricing power. Frontier labs are subsidizing customers today, charging below market cost to buy growth. When that growth must be shown as revenue and cash flow, they will need to raise prices — potentially at the worst moment — because open models are catching up fast. A Stanford-linked comparison shows open models closing a huge quality gap from early 2023 to near parity today, with overall model performance converging. When most models feel similar, competition collapses to price, and that crushes both revenue growth and margins.

Timing is therefore explicit: it is not yet. If Anthropic lists in October, its first reports would land around January or February; OpenAI could list even later, so the second and third prints that truly shape perception are nearly a year away. That delay leaves room for near-term catalysts to lift sentiment, notably solid Oracle results and IPO buzz around Anthropic that can keep AI excitement bid. In that window the video argues for using any September hole to stay on the hardware and infrastructure side of the train.

The valuation screen starts in infrastructure. On a forward non-GAAP growth-adjusted basis the host pegs Nvidia at 0.45 times, AMD around 1.0, Broadcom at 0.57, with TSMC, Arm, Marvell and Arista screening expensive at 1.2 to 3.0 times. The read-through is that within the same chip and networking family, Nvidia and Broadcom look like better bargains for the growth and profitability on offer, while AMD looks more than twice as pricey as Nvidia despite being its direct peer.

Memory and storage is where the money has actually been made. Over roughly three years SanDisk is up about forty-four-fold, Micron about fourteen-fold, Seagate about twenty-fold and Western Digital about ten-fold — even the laggard turned money ten times over. Year to date SanDisk leads at about 588%, followed by Micron near 241%, Seagate near 200% and Western Digital near 159%, with SanDisk still leading over the last month. The question is not who ran, but who remains attractively priced on a growth-adjusted basis.

Cybersecurity adds a different nuance. Revenue growth in the more mature management slice looks modest — Okta and Fortinet near 10-15% — yet profitability converts it into faster earnings. Palo Alto is pegged at 23% revenue growth against 23% earnings growth, CrowdStrike at 23% revenue versus 32% earnings, a leverage that suggests real platform pricing power and operating efficiency at CrowdStrike. Zscaler at 20% revenue versus 25% earnings growth stands out because the stock is down 42% over the past year despite that execution, flagging a disconnect between price and performance.

The rest of the basket fills the picture: Okta near 10-11% revenue with 17% forward EBITDA growth, Fortinet near 15% revenue with 16-17% earnings growth and a 24% realized pace last year that hints the guide may prove conservative, and Cloudflare near 30% revenue growth. The pattern the host likes is earnings outrunning sales, which signals that these firms prioritize profit as they scale, and the market does not always price it correctly. Fortinet's history of beating a modest guide is presented as a spot where surprise could create opportunity.

On a growth-adjusted price-to-earnings basis the stack reads Palo Alto about 4.2 times, CrowdStrike about 5.6, Zscaler about 1.6, Okta about 2.7, Fortinet about 2.7 and Cloudflare about 4.9. Waiting for cybersecurity to look cheap versus other sectors means waiting forever, so the approach is to pay up but pick the best value inside the group. The host flags Zscaler as the clearest bargain, CrowdStrike as expensive but worth selective accumulation, and Fortinet as a reliable, reasonably priced compounder with 2.7 times and steady profitable growth.

Practically, the strategy is two-layered: treat a September 10-20% slide as staggered heavy buying, especially in infrastructure and memory where demand still outstrips supply, and keep a disciplined calendar for the exit signal. The advice is to set news alerts the day OpenAI and Anthropic actually list, so each quarterly print — revenue and, just as importantly, profitability — is tracked in real time. The cautionary example is Micron's 14% plunge after a report that nearly tripled sales and more than doubled gross margin yet still disappointed the market's higher bar.

Stepping back, the big picture is nuanced rather than binary: near term choppy, September seasonally weak and valuations generous, but without audited profit evidence the AI story does not collapse. The longer threat is open-model convergence, commoditization and the need to turn subsidized growth into billed revenue before pricing power erodes. The calendar to mid-2027 offers flexibility; until then dips are framed as portfolio-expanding windows, afterwards discipline around cash flow and margins decides who remains a buyer.