Seventh Sense Blog

The $130B AI Capex Boom and What It Means for Email Marketers | Seventh Sense

Written by Mike Donnelly | Jul 15, 2026 2:00:00 PM

In the first quarter of 2026, the four largest hyperscalers — Google, Microsoft, Amazon, and Meta — spent more than $130 billion combined on data centers and AI infrastructure. In ninety days. The numbers are auditable in their SEC filings and grounded in the Bloomberg coverage of every earnings cycle. The headlines treat this as a tech-industry story. It is mostly a marketing-industry story, and the second-order effects on email are going to be larger and stranger than anything we've planned for.

Here's the macro-to-micro version: the AI infrastructure email impact isn't some abstract future thing. The compute being poured into the ground right now is going to land in your HubSpot portal over the next 18 months in concrete ways. Some of it will be great for marketers. Some of it will be quietly destructive. Most of it will be invisible until it isn't.

What $130B in AI capex actually buys

To make this concrete: the Q1 2026 capex from the four largest cloud providers is funding roughly three categories of buildout, all of which matter for email:

  • Inference capacity. The compute required to run trained models in production, at scale, cheaply. This is the part that affects per-send economics on AI features. As inference gets cheaper, things that were too expensive to ship in 2024 become standard in 2026.
  • Specialized model training. Models trained on narrow domains, including marketing-specific data sets that didn't exist commercially three years ago. Expect the next generation of email AI to be trained on email-shaped data, not general internet text.
  • Mailbox-provider AI. Don't forget the receiving side. Gmail, Outlook, and Apple Mail are all reinvesting heavily in inbox AI — filtering, summarization, prioritization. Their AI capex is going to reshape your delivery environment whether your vendor ships AI features or not.

That third bucket is the one most marketers haven't priced into their planning yet. Let's start there.

The mailbox is getting smarter faster than your sending stack

Here's an asymmetry worth sitting with: the AI infrastructure being built right now is being deployed on the receiving end before it's being deployed on the sending end. Mailbox providers have larger budgets, tighter user-facing feedback loops, and a clear business reason to invest. They're shipping inbox AI that does three things that materially change your job:

  • Summarization. Gmail's AI will summarize your three-paragraph nurture email into two sentences before the recipient ever reads it, as Google has documented across its Gmail AI announcements. Your subject line and pre-header now compete with an AI-generated synopsis of your own copy.
  • Prioritization. Inbox AI ranks senders by predicted relevance. If your sends look generic to the model, they get triaged into the "low priority" stack before a human sees them, regardless of whether they cleared the spam filter.
  • Reply suggestions. One-tap replies pull copy from your email. Whether that text reflects the action you wanted is now a function of how the model parsed your message.

The implication: email is no longer a direct conversation with a human. It's a conversation mediated by an AI that's deciding what to surface, summarize, and suggest. Writing for that mediated environment is a different craft than writing for an inbox in 2019 — and it's part of why open rates post-MPP have stopped meaning what marketers think they mean.

Second-order effects email marketers should plan for in 2026-2027

Working forward from the capex, here are the specific changes I'd expect to see hit B2B email marketing over the next eighteen months. Some of these are happening now. The rest are the trajectory.

1. Personalization gets dramatically cheaper

The dominant cost of personalization today is engineering and data, not compute. As inference gets cheaper, the math changes. Generating subject-line variants per recipient, scoring content blocks per persona, and dynamic body insertion all become economically defensible for mid-market senders, not just enterprise. Expect HubSpot, Marketo, and the rest to ship features that assume this cost curve.

2. Deliverability AI moves from add-on to baseline

The AI that watches your sender reputation, predicts engagement drops, and flags anomalies is currently sold as a premium add-on. With cheaper inference, this becomes table stakes — the way SPF/DKIM/DMARC checks for HubSpot sender reputation moved from "specialist feature" to "checkbox in setup" over five years. The vendors who don't ship it natively in 2027 will lose deals.

3. Generative content moves into the editor, not the autopilot

The autonomous-agent vision keeps getting deferred. What's actually shipping is generative AI inside the composer — better suggestions, smarter rewrites, faster localization. The capex makes these features fast and cheap enough to integrate into every editor by default. Expect this everywhere within twelve months.

4. The send-time category consolidates around recipient-level models

"Best time to send" tools that pick one window per audience are going to look as quaint as desktop-only email previews. Cheap inference makes per-recipient modeling the default, which is the architecture our scheduler has used since 2019 and is now being adopted across the category — with the measurable proof points for send-time optimization now widely documented. The Campaign Orchestration release we shipped in February 2026 is a preview of where the rest of the market is headed.

5. Mailbox-AI optimization becomes its own discipline

Just as SEO emerged when search engines started ranking pages, "mailbox AI optimization" is the discipline emerging now. Writing copy that summarizes well, structuring CTAs that survive prioritization, formatting emails that reply-suggestions render correctly — this is the next layer of craft. The marketers who get good at it early will have a window of asymmetric advantage.

What this means for your 2026-2027 budget

Three concrete budget moves worth making now, in priority order:

  1. Reallocate from generative tools toward infrastructure AI. The dollars spent on AI copywriters in 2025 should mostly go to deliverability, orchestration, and send-time AI in 2026. The ROI gap is widening fast.
  2. Train your team on writing for mediated inboxes. The bigger gain is not buying new AI. It's getting your existing copywriters to write for an environment where AI sits between you and the reader. This is a workshop, not a software purchase.
  3. Audit your stack for vendors who can't show their AI roadmap. Any martech vendor without a credible AI infrastructure investment is going to be priced out of competing on personalization, deliverability, and send-time by 2027. Their renewals are the line items most worth pressure-testing this year.

The contrarian take: most of the capex is wasted, and that's fine for you

Worth saying clearly: a meaningful fraction of the $130B in AI capex is going to be revealed, in retrospect, as overbuilt. That's how every infrastructure boom works — railroads, fiber, cloud, AI. Overbuilding is the mechanism by which the cost curve drops far enough for normal businesses to use the technology.

For email marketers, this is unambiguously good news. You don't have to fund the buildout. You get to use the depreciated infrastructure once the cycle prints capacity into the ground. The job is positioning your stack and your team to take advantage of it as the cost curve drops, not predicting which hyperscaler wins.

Frequently asked questions about the AI infrastructure email impact

Will AI capex make email cheaper to send?

The sending economics won't change much — SMTP delivery is already commoditized. What changes is the per-send economics of AI features layered on top of sending: personalization, send-time optimization, deliverability monitoring. Those drop in cost meaningfully as inference gets cheaper, which is how features that were enterprise-only move down-market.

How worried should I be about Gmail's AI summarizing my emails?

Concerned, but not panicked. Gmail's summarization is real and affecting engagement now, but the fix is craft, not technology: write tighter copy, lead with the action, and design CTAs that survive summarization. Marketers who adapt will outperform those who don't, and the adaptation isn't expensive.

Does the capex boom mean I should delay buying AI tools?

For generative tools, possibly — prices are dropping and features are improving fast enough that buying now means buying obsolescence. For infrastructure AI (deliverability, send-time, orchestration), buy now — the production track record matters more than the price curve, and the compounding benefit starts the day you turn it on.

Will hyperscaler AI features compete with marketing-specific AI vendors?

In some categories, yes. In email specifically, no — the domain expertise required to build deliverability-grade AI is not in the AWS or Azure playbook, and the data depth required is unique to vendors who've been in the space. Expect consolidation among marketing-AI vendors, not displacement by hyperscalers.

Where to go from here

The AI capex story is the kind of macro trend that's easy to ignore because it doesn't show up in your weekly metrics — until it does, all at once, when a mailbox provider ships a feature that reshapes your engagement curve overnight. The marketers who plan for the trajectory get to surf it. The ones who don't get washed.

If you want to see how recipient-level AI orchestration changes the math in your HubSpot portal, the free trial of Seventh Sense takes about 15 minutes to set up and starts modeling individual recipient behavior on day one. You'll see the kind of AI that's already crossed the gap from capex to results, in your own data.