The Trade Desk filed its FY2025 10-K on February 27. The stock was at $25, down 66% from its August peak of $91. Then on March 2-4, CEO Jeff Green bought 6 million shares for $148.1 million of his own money. The stock jumped 18% on the news.

The filing and the purchase tell opposite stories. This memo reconciles them.

What the 10-K Actually Says

The headline numbers confirm the bear case on volume:

Gross spend growth — the dollars advertisers push through TTD's platform — decelerated from +26% to +11%. Q1 2026 was guided at +10% revenue growth. The demand engine is slowing. This is what the stock priced in from $91 to $25, and the 10-K confirms it.

But look at the gap: revenue grew 18% while gross spend grew 11%. That 7 percentage point spread is the take rate expanding from 20.3% to 21.6%. The 10-K attributes this to "increased utilization of value-added services and data" and "increased pricing associated with value-added services and data." Translation: Kokai, their AI optimization platform, is letting TTD charge more per dollar of ad spend.

The Forensics: Is the Take Rate Real?

We checked every peer. The answer is unambiguous.

TTD is the only company in programmatic advertising expanding take rates. Every peer is flat or compressing. PUBM is actively competing on cost ("5x efficiency" from agentic campaigns). DoubleVerify's CEO confirmed global CPMs are declining.

TTD is charging more in a deflationary environment. That's real product differentiation, not industry tailwind.

The Paradox

Here's what nobody is talking about: TTD's take rate expansion makes the Amazon competitive threat worse, not better.

Amazon DSP charges a 1% take rate on open internet inventory. TTD now charges 21.6%. That gap was 19.3 points last year. Now it's 20.6 points. Every basis point TTD's take rate expands, the incentive for budget-constrained advertisers to test Amazon DSP increases.

Amazon isn't standing still. They've signed exclusive deals with Roku (80M CTV households), Netflix, Disney+/Hulu, Spotify, and SiriusXM. Jassy promoted the DSP in all three quarterly earnings calls in 2025. Wells Fargo surveyed TTD customers: 80% said Amazon DSP is already taking share of non-Amazon CTV budgets.

Green addressed this on the Q4 call. His framing: Google is the bigger competitor. Amazon DSP creates "channel conflict" with CPG brands (who compete against Amazon's retail business). Amazon is "mostly playing owned operated inventory."

He's technically right about today. Roku's CEO described Amazon DSP's open internet push as "early innings." But early innings become middle innings. And a 20.6-point take rate gap is the gravitational force accelerating that timeline.

This is the paradox: Kokai proves TTD's product works — it genuinely delivers better outcomes for advertisers (26% better CPA, 58% better CPR, 94% better CTR versus the prior platform). But the better the product, the higher the take rate, and the wider the pricing gap versus Amazon's loss-leader DSP. The bull case and the bear case are both getting stronger simultaneously.

The $148 Million Question

On March 2-4, Green bought 6 million shares through a limited partnership. We pulled the Form 4:

Transaction code P — open market purchase. Not an option exercise. Not a grant. He was the marginal buyer, moving the price up as he accumulated ($23.49 to $25.08 over three days).

This is one of the largest CEO open market purchases in any US public company in recent memory. And it happened while Green is personally named as a defendant in a Section 20A insider trading class action covering the period November 2023 to August 2025 — the exact window when the stock peaked and his CLO sold $4.7M at $91.

There are two readings:

Conviction: Green sees the FY2025 numbers, knows the Q1 pipeline, has visibility into Kokai adoption and the competitive landscape, and believes $25 is deeply undervalued. He's betting $148M that the growth deceleration is cyclical (CPG and auto soft), not structural (Amazon eating the franchise).

Litigation strategy: Green is named in an insider trading lawsuit. Buying $148M of stock at the trough creates a powerful legal narrative: "I'm so confident in this company that I'm putting $148M of my own money in." Securities defense lawyers know this playbook. The purchase creates reasonable doubt about whether prior sales were informed by inside information.

These readings aren't mutually exclusive. Green can believe the stock is cheap AND want to strengthen his legal position. The question is which is the primary driver.

The options market leans toward conviction. June 2026 options show call IV 5.7% above put IV — an inverted skew that's unusual. Normally the market prices more downside than upside risk. Here it's reversed. Max pain for June sits at $40, 34% above current. The positioning says: recovery, not further breakdown.

The Factor Decomposition

Regression (250 daily observations, multi-factor):

77.8% idiosyncratic variance. TTD moves on its own story, not the market. Highest in the ad-tech peer group (MGNI 56.8%, PUBM 68.2%, CRTO 37.2%). The drawdown was company-specific — growth deceleration and competitive threat, not sector rotation.

But high idio doesn't equal edge. TTD has 38 analysts covering it. The take rate expansion is in the 10-K that every one of them read. The Green purchase is a public Form 4 that hit every Bloomberg terminal instantly. The stock moved 18% the same day.

The five idio drivers:

1. Growth deceleration (bear, ≈40% weight) — Gross spend +26% to +11% to guided +10%. This IS the stock move.
2. Take rate / Kokai (bull, ≈25%) — Real, unique, corroborated. But in the filing.
3. Amazon DSP (bear, ≈20%) — 1% vs 21.6%. Early innings but trajectory clear.
4. CEO insider purchase (bull, ≈10%) — $148M. Extraordinary. Already priced.
5. Execution risk (bear, ≈5%) — No permanent CFO. Litigation active.

Forward alpha calculation: probability-weighted target $32.50, minus XLC sector return, gives ≈1.1% annualized idio alpha. With zero informational edge on any of the five drivers, adjusted alpha is zero.

What This Means

TTD is a genuinely interesting company in a genuinely interesting situation. The Kokai product works — the take rate data proves it, the peer comparison confirms it. Green's conviction is extraordinary. The valuation at 12.5x forward earnings on $1B operating cash flow with no debt is not demanding.

But interesting isn't the same as actionable. The five factors driving TTD are all well-understood by 38 professional analysts with management access, channel checks, and survey data. The filing confirmed what was priced. The Green purchase was extraordinary but priced in one trading day.

The honest assessment: TTD is in a doorway state where the bull case (Kokai monetization sustains growth despite volume deceleration) and the bear case (Amazon's 1% DSP slowly disintermediates at 20x lower cost) are both getting stronger. The resolution comes from Q2-Q3 2026 data: does gross spend reaccelerate, or does the deceleration continue into single digits?

Green is betting $148M that he knows the answer. He might. But that's his edge — visibility into the pipeline — not ours.

Predictions

Sources: TTD 10-K FY2025 (Feb 27, 2026), TTD Form 4 (Mar 4, 2026), MGNI 10-K/Q4 call, PUBM Q4 call, CRTO Q4 call, DV Q4 call, AMZN Q3-Q4 calls, ROKU Q4 call. All primary.