Javits Center, June 17, free admission, more than 145 sessions. AWS arrives in New York having announced end-of-support for its own Amazon Q Developer coding assistant while committing up to $25B more to Anthropic and 5 gigawatts of new compute, a clear bet on owning the model-and-infrastructure layer rather than the developer tool. Last year the summit was laser-focused on AI. This year the question is who pays for it.
Why this one matters
AWS Summit New York is AWS's flagship single-day East Coast gathering: cloud architects, platform and data engineering teams, enterprise CISOs, the FinOps and platform-ops crowd, and the dense partner ecosystem that sells observability, security, and data tooling on top of AWS. Jacob Javits Center, June 17, free admission, more than 145 expert-led sessions across data, agentic AI, security, and digital transformation. It matters this year because the agenda lands on top of AWS's loudest strategic signal in a while. In the last quarter AWS announced end-of-support for Amazon Q Developer, its own AI coding assistant, while committing up to another $25B to Anthropic and as much as 5 gigawatts of new compute to the partnership. Last year's NYC summit was, in one recap's words, laser-focused on AI. This year the room arrives knowing AI is the platform; the open question is who pays for it and who builds the layer that makes money.
The year's headline: AWS bets on the model layer, not the tool layer
The signal worth flying for is a retreat and a bet made in the same breath. (1) AWS announced end-of-support for Amazon Q Developer, the first-party AI coding assistant it launched into VS Code and the CLI, an unusual public step back from an application-layer product. (2) In the same window, Amazon committed up to another $25B to Anthropic, with Anthropic committing to spend over $100B on AWS compute over the decade and the two expanding collaboration toward 5 gigawatts of new capacity. (3) AWS launched frontier agents for security testing and cloud operations and brought Cerebras inference onto the platform. Read together, the strategy is clear: AWS is leaning into being the infrastructure-and-model host (the compute, the security agents, the Anthropic relationship) rather than trying to win the developer-tool layer itself. For an attendee, that reframes every booth and session, the question is no longer 'does AWS have an AI tool for this' but 'which partner fills the tool layer AWS just signaled it will not own.' The second beat, underneath, is cost. Forbes reported Uber burned its entire 2026 AI budget in four months on Claude Code, and the FinOps sessions exist because token economics are breaking enterprise finance assumptions in real time.
Sessions worth showing up for
The agenda lists 145-plus sessions and the keynote roster firmed up late, so target tracks and roles. (1) The opening keynote and any agentic-AI main-stage session: last year's NYC keynote put Intuit's Chief Data Officer on stage to walk through a production AI build, so watch which named customer AWS features this year, because the customer tells you which vertical AWS thinks is closing. (2) The security-and-governance sessions tied to AWS's new frontier agents for security testing: the clearest place to judge whether agentic security is shipping or still a demo. (3) The data and agentic-AI technical deep-dives: 145 sessions means the signal is in the 200-level rooms, not the keynote. (4) Any FinOps or cloud-cost-management session (Apptio brought a 'Shifting Left' cost session): in a year when a name-brand company exhausted its AI budget in four months, the cost-control rooms are where the most candid math gets spoken. (5) Favor anything tagged 'production' over 'getting started', the experimentation phase is over and the production sessions are where the real constraints surface.
Breakouts with signal density
A free single-day summit packs its signal into the partner-led technical sessions and the expo floor, not the keynote. Prioritize the observability and FinOps cluster, because that is where the cost reckoning is concrete: Coralogix (booth 921) is selling observability built for scale, Apptio is bringing a cloud-cost session, and the cost-data tooling category is quietly having its moment. The security sessions are the second cluster: HackerOne is on the floor pitching combined AI-and-human vulnerability testing, the human-in-the-loop counter-position to AWS's fully-agentic security testing pitch, and that contrast is the conversation. The data-platform partners (Denodo for integration, Smartling for content scale) are working the unglamorous 'your data has to be governed before the agents are useful' angle that every production AI story eventually hits. If you only have time for one room, make it a production case study with a named customer and a dollar figure attached.
Companies to track at the booths
Read the floor against the model-layer bet. AWS itself is the booth to read first: what it foregrounds (Bedrock and the Anthropic relationship, the frontier security agents, Cerebras inference) tells you where the platform is investing now that Q Developer is sunsetting. HackerOne says it combines advanced AI with human insight for vulnerability detection; what it is actually selling is the argument that fully-autonomous security testing still needs a human verdict, a direct hedge against AWS's own agent pitch. Coralogix says observability that is faster and built for scale; the real pitch is becoming the bill-control layer for teams whose AI and data volumes are now the line item finance is asking about. Apptio (an IBM company) says cloud financial management; what it is really selling at an AWS event in 2026 is FinOps as damage control after the first AI-budget overruns. Denodo says data integration for the AWS stack; the honest framing is that the agentic-AI demos do not work until the data layer underneath is governed, and that is the sale. Effectual, the 2025 AWS ProServe Rising Star partner hosting an off-site reception, is selling migration-and-modernization services, the consulting layer that AWS's first-party retreat just made more valuable.
Conversation patterns: three hallway debates, one nobody is saying
Debated in the hallway: (1) Whether sunsetting Amazon Q Developer means AWS cedes the AI-coding-tool layer to Claude Code and GitHub Copilot for good, or clears the deck for a Bedrock-native successor. (2) Whether the Anthropic compute commitment, up to 5 gigawatts, makes AWS customers more comfortable building on Claude or more nervous about a single-model dependency baked into the platform. (3) Whether the FinOps panic is a 2026 spike that tooling will absorb, or the permanent new shape of cloud budgets now that inference is a variable cost tied to usage. The thing nobody will say from the stage: a free single-day summit with 145 sessions is a developer-and-architect retention event, and it is running at the exact moment enterprises are auditing what they spend on AWS-hosted AI, so the unstated job of the day is to keep the platform sticky while the bill is under review. The vendors who name the cost problem out loud are doing the customer's work for them; the ones who route around it are hoping you renew before the FinOps review lands.
After the badges come off
A free Javits summit is a high-density, one-day firehose: 145 sessions, a packed expo floor, and the partner receptions stacked into the evening (Effectual's reception at Spygold is one of several). You leave with a badge full of scans, a stack of cards from AWS solutions architects, partner reps, and the observability-security-data vendors who all blurred together by 4pm, and an inbox that resets to your day job on Thursday morning. Every one of those follow-ups competes with the same person's stack from the same floor. The conversation at the Coralogix booth or in the FinOps session is worth exactly what the follow-up that lands before the week ends is worth. Met captures the context while you still remember which conversation was the one worth continuing, so your follow-up reads like the conversation kept going, not 'great to meet you at the AWS Summit.'
Download Met for iPhone
Read by cloud architects, platform and data engineers, and the security and FinOps crowd heading to Javits for AWS Summit NYC, pre-event analysis pulled from AWS's last 90 days of public signal, the Q Developer sunset, the Anthropic compute deal, and the enterprise AI-cost cycle. No signup, nothing stored on our servers.
