Meta Just Poached AWS's EC2 Chief. The Follow-On Pool Moves in 90 Days.

On July 18, 2026, WSJ confirmed what CNBC had trailed three days earlier: Dave Brown, a 19-year AWS senior vice president and S-Team member running EC2, Lambda, Bedrock, and SageMaker, is leaving Amazon at the end of July to build "Meta Compute." If you are a sourcer or a founder trying to hire AI infrastructure principals, the exec is not your target. The 30 to 60 people who quietly follow him over the next 90 days are.

That is the pool worth building a list around this quarter, and it is smaller and more geographically clustered than most recruiters expect.

Why the Brown hire is a sourcing signal, not a leadership story

The Dave Brown to Meta move is a leading indicator that a specific, nameable set of AWS compute and ML principals will be recruitable inside a 90-day window. The exec headline is already spoken for. The follow-on trail is not.

Brown will report to Santosh Janardhan, Meta's head of infrastructure, and joins as Zuckerberg's "Meta Compute" initiative (introduced in January) tries to become a fourth hyperscaler backed by $125B to $145B of 2026 capex. His AWS successor, Dave Treadwell, takes over compute and ML on August 1. Treadwell is a 27-year Microsoft veteran who has run Amazon's eCommerce Foundation since 2016. He is not a cloud-native successor, and that detail is the whole story for the lieutenants underneath.

Here is why every AI infrastructure recruiter should treat the next 90 days as a hard sourcing window:

• Brown was promoted to SVP only in April 2026. A three-month gap between promotion and exit is unusually short and signals a pre-planned build.
• AWS just posted 28% revenue growth. This is a departure from strength, which lowers the stigma for lieutenants who follow.
• Meta already ran this play in 2025 with Apple. After Frank Chu joined Meta's Superintelligence Lab, at least six Apple AI engineers followed within roughly 60 days.
• Amazon non-solicits typically bind for ~12 months on employee lists. Brown's personal knowledge of who to call is not contractually bindable.

How big the follow-on pool actually is

The addressable follow-on pool for Meta Compute is roughly 30 to 60 principal engineers, staff PMs, and infra directors over 90 days, concentrated in three cities. The number is small because Brown's real network is not "AWS," it is the 2007 to 2012 EC2 founding cohort.

Brown joined Amazon in 2007 in Cape Town, South Africa, where the original EC2 engineering team was seated. Any recruiter filtering for "AWS + Seattle" will miss the highest-conversion candidates. The Cape Town diaspora, the early Herndon (US-East-1) operators, and a small Seattle senior bench are where the deepest loyalty sits.

28

The table below is the dataset the estimate is built on. Treat it as the skeleton of a target list, not a final count.

The Chu precedent is the cleanest comparable, and the ratio (one senior hire, six documented IC follows in 60 days) is the anchor. Brown's org is roughly five times the size of Chu's Apple sphere of influence, but his AWS network has more constraints (RSU cliffs, non-solicits, longer tenure). Net it out and you land in the 30 to 60 range.

The three cities that matter, and the one that doesn't

The Brown follow-on trail concentrates in Cape Town, Herndon, and Seattle. It does not concentrate in the Bay Area, which is where most Meta-adjacent sourcing lists start.

• Cape Town. The original EC2 engineering hub. Brown's 2007-era colleagues are still there. LinkedIn coverage is thin, GitHub coverage is patchy, and most US recruiters do not source into South Africa.
• Herndon, Virginia. The US-East-1 operations bench and a large fraction of the AWS compute PMs. Under-recruited because the surrounding market is government cloud, not FAANG.
• Seattle. Obvious, but the useful signal here is not "worked at AWS." It is "worked in EC2 or Lambda before 2015." That filter cuts the noise by an order of magnitude.
• Not the Bay Area. Meta's Menlo Park recruiters will overweight local candidates. The Brown-loyalist pool is not local.

This is the exact gap Refolk closes for hyperscaler talent poaching: you describe the person in plain English ("staff or principal engineers who worked on EC2 or Lambda between 2008 and 2014, based in Cape Town, Herndon, or Seattle") and get a ranked shortlist across GitHub, LinkedIn, and the open web, without pre-building a Boolean.

Why Treadwell's arrival is the real accelerant

Dave Treadwell taking Brown's seat on August 1 is the push factor that turns a normal exec exit into a lieutenant exodus. When an S-Team compute seat passes to a non-cloud-native successor, the compute and ML principals lose their internal sponsor.

Treadwell spent 27 years at Microsoft before joining Amazon, and has run eCommerce Foundation (retail infrastructure, not cloud) since 2016. That is a fine resume. It is not the resume of someone the Bedrock or SageMaker principals will follow into an internal roadmap fight. The mechanics that follow are predictable:

1. Day 0 to 30. Brown's direct reports get "stay" packages. Roughly a third accept without conditions. The rest wait.
2. Day 30 to 60. Q3 RSU vesting hits. Principals who were waiting on cliffs make decisions. The first tranche of moves gets announced on LinkedIn.
3. Day 60 to 90. Meta Compute's org chart leaks. Principals who did not get scope in the new AWS structure surface on the market. This is the largest tranche.
4. Day 90+. Non-solicit clocks are still running, but personal networks are not gated by them. The remaining moves happen through warm intros, not inbound.

The 90-day window is real and mechanical, not cultural. Vesting decisions crystallize in tranches, and so do the moves.

Compare this to the 2024 Adam Selipsky exit, which chilled follow-on movement because it read as instability. Brown is leaving during a 28% growth quarter. There is no rat-leaving-the-ship story to tell, which means the psychological cost of following him is close to zero.

The 2026 AWS exit pattern this fits into

Brown's move is not isolated. It is the fourth major AWS exit in the first seven months of 2026, and the pattern points at AI infrastructure recruiting getting hotter, not calmer.

• David Luan, Head of AWS's AGI Lab and Agentic AI R&D, exited earlier in 2026.
• Scott Rosecrans, VP of AI Sales, left AWS for OpenAI.
• Jon Jones, a longtime AWS VP, departed in 2025 amid a wave of eight documented VP+ exits.
• Dave Brown, SVP compute and ML, exits end of July 2026 for Meta Compute.

Every one of those exits carries its own follow-on trail. Luan's went to agentic AI startups. Rosecrans's went to OpenAI's enterprise motion. Brown's will go to Meta. If you are sourcing for AI infrastructure right now, the useful move is to keep a live list of the S-Team and the layer beneath it, and to check for movement weekly. That is the sort of standing query Refolk is built for: "principal or staff engineers who reported to Dave Brown at AWS in the last five years, sorted by recency of LinkedIn activity."

What Meta's offer actually looks like (and why it isn't cash)

Meta's differentiator for the Brown follow-on pool will be scope, not compensation. This runs against the 2025 narrative and matters for how recruiters position competing offers.

In 2025, Meta paid "tens of millions" for individual AI researchers from OpenAI, Google DeepMind, Anthropic, and Apple. That worked because the target pool was 200 people worldwide and the roles were research. Infrastructure principals are a different market. There are more of them, they are more comp-anchored to public-company bands, and they optimize for a different thing: greenfield.

Meta Compute is the only greenfield hyperscaler build available in 2026. AWS is mature. Google Cloud is mature. Azure is mature. The Louisiana Hyperion project alone is expanding from 2 GW to 5 GW with more than $50B of investment. If you are a Lambda principal who wants to build compute from the ground up again, you cannot get that at Amazon anymore. That is the offer.

$145B

For AWS EC2 engineer sourcing right now, this means three things:

• Do not lead with cash. The Brown-loyalist pool is not underpaid at AWS.
• Lead with scope specificity. "Build the L2 network fabric for a 5 GW campus" beats a comp number.
• Time the outreach to the vesting cliffs. August, September, and October each have distinct AWS RSU dates. The message that lands on the wrong week gets ignored.

The two-week playbook for sourcers

If you are trying to get in front of the Brown follow-on trail before Meta's in-house recruiters lock it up, the window is now, and the playbook is four moves.

1. Build the seed list this week. Start with anyone who reported to Brown between 2015 and 2026, plus anyone with EC2 or Lambda in their title before 2015. Cape Town, Herndon, Seattle. Ignore the Bay Area for now.
2. Layer the Chu precedent. Map the six-in-sixty-days Apple pattern onto Brown's org. Who are the six most likely first-tranche movers? Message them before day 30.
3. Monitor for the second tranche. Set a weekly check on the layer below the S-Team. LinkedIn "Open to Work" is a lagging signal. GitHub activity drop-offs, updated bios, and new "advising" roles are leading ones.
4. Do not chase Brown himself. He is not moving again for at least three years. His direct reports are, and their direct reports are, and that is the pool.

For anyone running this playbook against a real req, Refolk collapses the seed-list step from a week of Boolean into a paragraph of English, and keeps the standing query live so the second and third tranches surface without a re-source. That is the sourcing muscle that matters when the follow-on pool is 47, or 30, or 60. Small enough to work by hand, too small to miss.

FAQ

Who is Dave Brown and why does his move to Meta matter?

Dave Brown is a 19-year AWS veteran and S-Team member who ran EC2, Lambda, Bedrock, and SageMaker. On July 18, 2026, WSJ reported he is leaving AWS at the end of July to build "Meta Compute," Zuckerberg's initiative to make Meta a fourth hyperscaler backed by up to $145B of 2026 capex. His move matters less as an exec story and more because SVP-level hyperscaler exits reliably drag a 30 to 60 person follow-on trail of principals within 90 days, and that trail is the sourceable pool.

How big is the follow-on pool for Meta Compute, actually?

The defensible estimate is 30 to 60 principal engineers, staff PMs, and infra directors over 90 days, based on the Frank Chu precedent (one senior Apple hire, six documented IC follows in 60 days) extrapolated to Brown's larger AWS org, with AWS RSU cliffs and non-solicits factored in as friction. The pool concentrates in Cape Town, Herndon, and Seattle, not the Bay Area, because Brown's founding EC2 network is geographic rather than org-chart-based.

When is the sourcing window and why 90 days?

The window is roughly August through October 2026, tied to Q3 AWS RSU vesting dates and the August 1 arrival of Dave Treadwell as Brown's successor. The 90 days is mechanical: initial "stay" packages get accepted in the first 30 days, vesting-cliff decisions crystallize between days 30 and 60, and the largest tranche moves at 60 to 90 days once Meta Compute's org chart leaks and internal roadmap fights at AWS resolve.

How should recruiters position Meta Compute against a stay-at-AWS counter?

Lead with scope, not cash. Meta paid tens of millions for individual AI researchers in 2025, but infrastructure principals are more comp-anchored to public-company bands and less willing to be bought. The real lever is that Meta Compute is the only greenfield hyperscaler build available in 2026, with the Louisiana Hyperion campus expanding from 2 GW to 5 GW. Any Lambda or EC2 principal who wants to build compute from scratch again cannot get that at Amazon anymore.