Mizuho Reverses SaaS Collapse: Figma's 46% AI Growth Shocks Markets

So, What Actually Happened?

Wednesday morning the strategy desks are reading a Tuesday that handed enterprise teams four named operating shifts at once. The ”SaaS Rout of 2026” just reversed after Mizuho channel checks showed enterprise buyers still investing in data modernization, with Figma posting 46% revenue growth on AI monetization. We scanned 190,000 articles this week so you don't have to. Meanwhile Brussels rewrote the EU AI Act calendar, giving healthcare AI 16 extra months while pulling nudifier-app prohibitions forward, and NextEra and Dominion signaled the era of AI utility mega-mergers has officially started.

The Bottom Line: When the consensus ”AI is killing SaaS” thesis collapses on a Tuesday channel-check, when Brussels gives healthcare AI 16 more months but tightens the rest of the AI Act, and when two of the largest US utilities start consolidating to feed AI power demand, the operating-floor question shifts in the same week: not whether to bet on AI, but which regulated layer (compliance calendar, energy supply chain, agentic workflow ownership) you have a named owner for. The CIO walking into Thursday with one page mapping the SaaS-vs-agent question, the EU calendar, the energy procurement story, and the agent-distribution layer runs the next quarter. The rest will be reading from a March deck.

The Tracks That Matter

1. The 'SaaS Rout Of 2026' Just Reversed In A Single Channel Check

The sharpest valuation signal of the week sits on a Mizuho channel-check note that reversed the SaaS thesis most Wall Street desks had baked into Q1 models. Enterprise buyers, the sell-side found, are still investing in data modernization, still expanding AI-tool budgets inside existing SaaS contracts, and still showing growing interest in newer AI features rather than ripping out the platform underneath. Figma posted 46% revenue growth with real AI monetization traction. ServiceNow announced a multi-year AI partnership with Experian. The ”AI will eat SaaS” framing that anchored most 2025 SaaS-stock thesis decks just got a named operating-floor counter-reference.

The contrast that sharpens the read is what most enterprise software CFOs had pencilled into 2026 budget conversations: a defensive crouch, a renewal cycle dominated by ”show me the AI angle or we are cutting seats,” and a procurement narrative built around hyperscaler-native model studios replacing third-party platforms. After a single sell-side channel-check showed enterprise relationships, proprietary data, and workflow integration are still doing the load-bearing work, the conversation flips. The platform with the data plus the workflow plus the embedded AI feature wins the renewal, not the standalone AI feature without the workflow.

The strategic implication: the CFO and head of platform just gained a named ”AI-feature-inside-SaaS vs. AI-feature-replacing-SaaS” question to put on the next architecture review. The right answer probably runs to the SaaS incumbent for the next two cycles, which most aggressive AI-transition decks did not assume.

Here's what works: Ask the CIO and CFO together: for our top three enterprise SaaS contracts up for renewal in the next two quarters, do we have a documented ”AI feature roadmap inside this contract” line item, or are we still planning around an AI-feature replacement we have not actually scoped? Mizuho just priced the optionality question.

2. Brussels Just Rewrote The AI Act Compliance Calendar

The most under-reported regulatory signal of the week sits on a TechTimes brief on the EU's AI Act timeline rewrite. Healthcare-AI vendors get 16 more months on the high-risk-system compliance window. Hiring-AI vendors get an extended runway. New prohibitions on nudifier apps and certain manipulative-AI categories were pulled forward, not pushed back. The InsidePrivacy analysis from Covington confirms the same shape: timeline relief in the regulated-vertical layer, targeted simplification in the obligations stack, fresh prohibitions on the consumer-harm layer.

The contrast is what most legal-tech and procurement teams still have on the GRC calendar: a single August 2026 mental model for ”the EU AI Act bites.” After the rewrite, the calendar is at least three different dates: the new healthcare/hiring-vendor extension, the unchanged general-purpose model obligations, the pulled-forward prohibition list. The ”we are ready for the AI Act because we have the high-risk-system register” framing that anchored most 2025 readiness assessments just got reset.

The strategic implication: the GC, chief privacy officer, and head of platform just gained a named ”version-2 EU AI Act calendar” row on the GRC scorecard. The vendor selection question is no longer ”does this vendor pass the August 2026 bar.” It is ”which sub-deadline does each of our regulated-AI workflows now sit on, and who owns the named filing for that specific window.”

Here's what works: Ask the GC and chief privacy officer together: for our top three regulated-AI workloads (hiring, healthcare, consumer-facing, financial decisioning), do we have a named owner per sub-deadline on the new EU AI Act calendar, with a filing checklist tied to the specific date that applies to that workload? Brussels just turned one date into four.

---

Try It Yourself

---

3. NextEra And Dominion Signal The Era Of AI Utility Mega-Mergers

The clearest infrastructure signal of the week sits on a Financial Post brief on the NextEra-Dominion deal. Two of the largest regulated US utilities just signaled a consolidation move framed explicitly around AI power demand and data center capacity. The pairing matters because regulated utilities have not done mega-mergers in two decades, and when they move, they move because the underlying load profile changed. The Cooley alert on financing the AI energy boom lands the same week and tells the financing side of the story: data, power, and dollars are now one financing conversation, not three.

I have been DJing the data warehouse long enough to recognize this sound. It is the rumble of the bass bin a floor below the booth, the one that tells you the kick drum about to drop is going to move the whole room. Two utilities consolidating do not announce themselves as a tech story. The room reads them as a tech story by the second hour. The ”AI capacity is a hyperscaler conversation” framing already retired the procurement seat. The ”AI capacity is also a utility-regulator conversation” framing just took the floor.

The strategic implication: the CFO, head of platform, and chief sustainability officer just gained a named ”utility consolidation watch” row on the procurement scorecard. The question is no longer only ”do we have a non-hyperscaler GPU contract.” It is ”do we have a named utility-side line on the power supply behind the GPU contract, and is that utility still the same one we modeled twelve months ago.”

Here's what works: Ask the head of infrastructure and the head of sustainability together: for our top three AI-heavy workloads, do we have a named utility-side power supplier on the file with a contracted demand window and a regulator-jurisdiction note, or are we still treating power as an undifferentiated line on the cloud bill? NextEra just made the utility a procurement signal.

4. Coatue's $12T AI Bet Just Split The Tech Thesis

The most uncomfortable capital-markets signal of the week sits on Coatue's May 2026 AI report breakdown. The hedge fund's research note pegs the AI buildout at $12 trillion in cumulative spend through 2030 and splits the tech investor universe into ”believers in the buildout” and ”skeptics on returns.” Neither side gets a victory lap in the report. The buildout side cannot show defensible unit economics on the inference layer yet. The skeptic side cannot point to the consumer-AI revenue collapse that would prove the bubble call. The note reads as a recalibration, not a recommendation.

The contrast that sharpens the read is what most enterprise CFOs still see in their CapEx model: a single line item for ”AI infrastructure” priced against a single ROI assumption. After a top-tier hedge fund publishes a $12T cumulative-spend scenario without endorsing it as upside, the assumption underneath every ”AI is the next cloud” CFO pitch deck needs a second look. The edithyeung weekly tally of 19 Silicon Valley startups raising $770.4M last week sits in the same news cycle: capital is still flowing into the buildout, but the spread between ”raised at high valuation” and ”delivering durable revenue” is widening.

Here's what works: Ask the CFO and head of strategy together: for our 2026-2027 AI CapEx, is the underlying ROI assumption documented as a number with a named owner and a quarterly review trigger, or are we running on the same ”AI as the next cloud” pitch deck the rest of the market is using? Coatue just put a $12T scenario on the table without endorsing it. That is the framing your audit committee will be asking about by Q4.

5. UiPath Just Redefined What 'Software Builder' Means

The sharpest workflow-ownership signal of the week sits on a Techzine piece on the UiPath CEO interview. Daniel Dines reframed agentic AI not as a feature on top of the RPA stack but as a redefinition of who counts as a software builder inside the enterprise. The retail-investor read on the same shift called it a bet-the-company pivot. The core claim: the line between ”business analyst describing a workflow” and ”engineer shipping an automation” just collapsed into one role, and the platform that owns that role owns the next decade of enterprise productivity software.

The contrast that sharpens the read is what most CIOs still have on the org chart: a development team, an analyst team, a process-automation team, and a fourth team trying to do agent pilots without authority to ship. The ”agentic AI is a feature we are evaluating” framing that anchored most 2025 IT-strategy decks just got a named counter-reference from the vendor that has been quietly owning the end-to-end workflow layer in the Fortune 500 for a decade.

Here's what works: Ask the CIO and head of platform together: for our top three high-volume workflows currently split across analyst, developer, and automation teams, do we have a named ”single workflow owner with shipping authority” candidate on the org chart, or are we still routing every agent proposal through a three-team approval loop? UiPath just made the org-chart question a vendor-selection question.

6. Data Governance Just Became The New Security Perimeter

The quietest security signal of the week sits on a Logicalis brief framing data governance as the new perimeter for agentic AI deployments. The argument: once an autonomous agent can read, summarize, and act on enterprise data, the network firewall becomes a footnote. The real control plane is the data-access layer the agent reads from. The Black Duck Polaris May 2026 platform update lands the same week with explicit AI-security tooling for SCA and SAST workflows. Two independent vendor signals, one shared frame: the security perimeter is moving from the network to the data plane.

The contrast is what most CISO scorecards still measure: endpoint coverage, EDR alerts, network-traffic anomalies, identity-provider events. After a regulated agent reads a customer record and acts on it, the failure mode is not a network breach. It is the data-governance contract behind the read. The ”we are ready for agentic AI because we have endpoint coverage” framing that anchored most 2025 security-budget proposals just got a named counter-reference.

Here's what works: Ask the CISO and chief data officer together: for our top three agentic AI use cases moving toward production, is the named control plane a network firewall plus an identity check, or is it a data-governance contract per data domain with a named owner and an audit trail? Logicalis just put the perimeter question on the wrong slide.

7. OTB Ventures Just Booked The Quietest Industrial-AI Series B Of The Week

The most under-covered funding signal of the week sits on a Greenberg Traurig note on the OTB Ventures-led Series B in SPREAD AI. SPREAD AI builds an enterprise engineering platform that combines knowledge graphs and AI for industrial product development, the kind of category that does not get a TechCrunch homepage but lands inside the engineering function of every automotive, aerospace, and heavy-industry customer. The edithyeung tally of 19 Silicon Valley startups raising $770.4M shows the same week's frothier consumer-AI rounds in San Francisco. SPREAD is the contrast: industrial engineering, European cap table, knowledge-graph backbone.

The contrast that sharpens the read is where every ”AI is hitting the enterprise” deck still points: customer service, marketing copy, sales-call summarization. The actual margin-protection layer for industrial firms is the engineering-knowledge function: bills of materials, design history, supplier rationalization, change orders. When a Series B closes on an industrial-AI platform built on knowledge graphs (not LLM wrappers), the European industrial buyer just gained a named ”AI inside the engineering function” reference, and the US consumer-AI narrative is missing it.

Here's what works: Ask the CTO and chief engineer together: for our top three engineering-knowledge bottlenecks (BOM rationalization, design history search, supplier-change impact analysis), do we have a named knowledge-graph-backed AI platform on the evaluation list, or are we still piloting LLM wrappers on top of a SharePoint folder? SPREAD just priced what a real industrial-AI Series B looks like.

Signal vs. Noise

🟢 Signal: Regulatory Compliance. Regulatory compliance climbed in real influence on the wires Tuesday into Wednesday morning while the generic ”AI” label kept growing in volume but lost ground in named operating reach. Enterprise buyers are quietly moving the AI conversation from product features down to the compliance-calendar layer (EU AI Act sub-deadlines, healthcare AI extension, CCPA cybersecurity audits), the same one most coverage filed under ”boring.”

🔴 Noise: Machine Learning As A Single Topic. ”Machine learning” pulled the most mentions across the wires this morning but its real operating influence dropped, because the conversation already split into named sub-tracks (agentic workflow ownership, AI-utility power consolidation, data-governance security perimeter, EU AI Act sub-deadlines). Anyone still tracking ”machine learning” as a single signal is reading from a 2023 keyword filter while the operating floor split it into a dozen named files.

From the 190K

We scanned 190,000 articles this week. Here's what no one's talking about:

The SaaS-rout reversal at Mizuho, the EU AI Act timeline rewrite in Brussels, and two regulated US utilities signaling an AI-driven mega-merger all printed inside one 48-hour window.

Each desk reads these as unrelated stories. The capital-markets desks file Mizuho. The privacy-and-compliance desks pick up the EU rewrite. The energy press writes the utility deal. Read them on the same morning and a different shape emerges: the AI conversation just got named compliance dates, named energy suppliers, and named platform incumbents in the same news cycle. The ”AI is a single procurement conversation we are running this year” framing that anchored most mid-market 2025 strategy decks just got three named counter-references at once, each from a different layer of the operating stack (vendor renewal, regulator filing, power contract).

The strategic move on Thursday is mapping which of those three operating layers (named incumbent SaaS owner, named EU sub-deadline owner, named utility-side power owner) has a real owner inside your own stack, and which one still says ”TBD.” The week just priced the gap.

By The Numbers

• Figma posted 46% revenue growth with AI monetization showing genuine traction, while ServiceNow announced a multi-year AI partnership with Experian — Two named enterprise-SaaS incumbents are growing on AI features inside the contract, not despite the AI threat. The SaaS-versus-AI thesis just got two named operating counter-references.

• Coatue's May 2026 AI report puts cumulative AI buildout spend at $12 trillion through 2030 — A top-tier hedge fund's research note, published without an upside endorsement, on the size of the AI capital cycle. The investor universe is now split between ”believers in the $12T number” and ”skeptics on the returns.”

• 19 Silicon Valley startups raised $770.4 million in a single week — A weekly tally from a venture investor's newsletter. AI-adjacent capital continues to flow at peak velocity into the SF Bay even as Coatue's report flags the durability question.

• Healthcare AI vendors received 16 additional months on the EU AI Act high-risk compliance window — The Brussels regulator rewrote the compliance calendar with a sector-specific delay, while pulling forward prohibitions on consumer-harm AI categories.

• The AI in drug discovery market is projected to reach $6.89 billion by 2029 at a 29.9% CAGR — A pharma-vertical sizing read from a sector tracker. The ”AI in pharma is hype” framing keeps getting lapped by named platform adoption inside the R&D function.

• Maine public universities reached a $1.39M deal for an AI tool agreement — A regional-public-sector AI procurement number that lands inside the same window as the EU compliance rewrite. The public-sector AI spend pattern is moving from federal-only to state-by-state line items.

• Globant launched a $125 million share repurchase program — A digital-services firm doing a buyback while AI-implementation revenue is rising. Capital allocation inside services firms is shifting toward shareholder return while the AI-implementation revenue line continues to ramp.

• See what's rising across the 190K-article corpus this week →

Deep Dive: The Strategy Map Just Grew A New Layer

Every DJ knows the moment when the headline track ends and the room expects silence, but the back-catalogue cut underneath has been holding the energy the whole time. The crowd notices the new track only when the headline ends. This week, the enterprise-AI room had three back-catalogue cuts pulling underneath the model-lab funding theatre: a SaaS-rout reversal, an EU compliance calendar rewrite, and a utility-side power consolidation. None of them got the homepage placement. All three changed the strategy map.

The Incumbency Track

The Mizuho channel-check reversal is the capital-markets equivalent of saying: the entrenched SaaS platform with the data, the workflow, and the embedded AI feature is winning the renewal cycle the AI-native challenger was supposed to win. Figma's 46% growth on early AI monetization is the punch line. The ”AI-native challenger replaces incumbent SaaS in the renewal window” framing that anchored most 2025 transition decks just got a real counter-reference from inside the procurement function. The CFO who walks into the next architecture review with a defensive AI-feature-replacement plan instead of an AI-feature-inside-incumbent plan is going to spend the rest of the year explaining why the renewal got cheaper and the data got harder to move.

The Compliance Calendar Track

The EU AI Act timeline rewrite is the regulatory equivalent of saying: the single August 2026 date most legal teams pencilled in just split into four sub-deadlines, with different filings for different verticals, different obligations for different model classes, and different prohibition timelines for different consumer-AI categories. The GC who walks into the next risk review with a one-date AI Act mental model is the one explaining to the audit committee why the healthcare-AI procurement that should have closed in Q2 is now waiting on a vendor that did not have the right sub-deadline owner on the contract.

The Energy Supply Track

The NextEra-Dominion utility consolidation is the infrastructure equivalent of saying: AI compute now has a regulated-utility procurement dependency that did not exist 18 months ago. The CFO who has only modeled cloud-cost growth in the AI CapEx line is missing the second-order question: which regulated utility is on the file behind the cloud region, and is that utility about to consolidate into a different rate base. The Cooley financing-the-AI-energy-boom read makes the financing layer explicit: data, power, and dollars are now one underwriting conversation, not three.

What Actually Works

1. Put a named AI-feature-inside-incumbent row on every SaaS renewal scorecard. Every renewal cycle gets a named ”AI feature roadmap inside this contract” line item before the AI-feature-replacement conversation. Mizuho just priced the incumbency advantage for you.

2. Put a named EU AI Act sub-deadline owner on every regulated AI workflow. Every workflow gets a named owner per sub-deadline (healthcare, hiring, general-purpose, consumer prohibition), with a filing checklist tied to that specific date. Brussels just turned one date into four.

3. Put a named utility-side power supplier on every AI architecture review. Every AI-heavy workload gets a named utility-side row alongside the GPU-vendor row, with a regulator-jurisdiction note and a consolidation-watch flag. NextEra just made the utility a procurement signal.

4. Put a named data-governance contract on every agent deployment. Every agentic AI workflow gets a named data-governance contract per data domain, with an audit trail and an owner. Logicalis just moved the security perimeter from the network to the data plane.

The model-lab funding rounds will come back. They always do. The next benchmark, the next safety paper, the next valuation rumor will print by Friday. The room is still moving. The operator who walks into Thursday with the incumbent-renewal row, the EU sub-deadline owner, the utility-side power line, and the data-governance contract already on the dashboard is the one mixing for the rest of the quarter. The one who waits for the headline track is going to play to a thinner floor by Q3.

What's Coming

The First Top-50 Enterprise To Disclose Multiple EU AI Act Sub-Deadline Owners In One Filing

The EU AI Act timeline rewrite is the trigger. The next move is the first top-50 multinational disclosing, in an annual report or risk filing, a named owner for each EU AI Act sub-deadline (healthcare, hiring, general-purpose model, consumer prohibition) with a sub-filing checklist. That disclosure is probably one to two cycles out, and the GCs who already mapped the calendar absorb it as routine.

The First Public Disclosure Of A Named Utility-Side Counterparty In An AI CapEx Note

The NextEra-Dominion deal is the trigger. The next move is the first AI-heavy enterprise (cloud-native, retail, financial services) disclosing a named regulated-utility counterparty in a CapEx footnote, with a contracted-demand window and a rate-base jurisdiction. That disclosure is probably one to two cycles out, and the CFOs who already pre-staged the utility-watch line absorb it as routine.

The First Named Industrial-AI Platform In A Fortune 500 Engineering Function

SPREAD AI's Series B is the trigger. The next move is the first Fortune 500 industrial firm (automotive, aerospace, heavy industry) disclosing a named industrial-AI platform inside the engineering function, with a BOM-search or design-history use case live in production. That disclosure is probably two to four cycles out, and the CTOs who already piloted the knowledge-graph backbone are the ones running the case study.

For Your Team

Strategic purpose: Thursday is the day this week's incumbency-reversal, EU-calendar-rewrite, utility-consolidation, and industrial-AI signals get translated into one Strategy Map before the next architecture or risk review. The work is one named owner per load-bearing layer: the SaaS-incumbent renewal row, the EU sub-deadline owner, the utility-side power line, and the data-governance contract per agent. Everything else is commentary.

Thursday's meeting prompt: ”If Mizuho just reversed the SaaS-rout thesis Tuesday morning, if Brussels just turned one EU AI Act deadline into four, if NextEra and Dominion just signaled the era of AI utility mega-mergers, and if Logicalis just argued the security perimeter is now the data-governance contract, who in this room owns the named scorecard across our SaaS-renewal layer, our EU compliance calendar, our utility-side power layer, and our data-governance layer, and is that owner one person, or four people who have never been in the same meeting?”

The Strategy Map Framework:

1. One named owner per load-bearing AI layer. CIO and CFO co-own the SaaS-incumbent renewal row. GC and chief privacy officer co-own the EU sub-deadline calendar. Head of infrastructure and chief sustainability officer co-own the utility-side power line. CISO and chief data officer co-own the data-governance contract per agent.

2. Named AI-feature roadmap inside every incumbent SaaS contract. Every renewal cycle gets a named AI-feature roadmap line item before the AI-feature replacement conversation. Mizuho priced the incumbency option for you.

3. Named owner per EU AI Act sub-deadline. Every regulated-AI workflow gets a named owner per applicable sub-deadline, with a filing checklist tied to the specific date. Brussels turned one date into four.

4. Named utility-side power supplier per AI workload. Every AI-heavy workload gets a named regulated-utility counterparty with a contracted-demand window and a consolidation-watch flag. NextEra made the utility a procurement signal.

5. Named data-governance contract per agent deployment. Every agentic AI workflow gets a named data-governance contract per data domain, with an audit trail and an owner. Logicalis moved the perimeter to the data plane.

Share-worthy stat: A single Mizuho channel-check just reversed the SaaS-versus-AI thesis after Figma posted 46% revenue growth on AI monetization. Drop that on the next architecture review and the ”AI-native challenger replaces incumbent SaaS” framing reframes itself in 30 seconds.

Go deeper: Track the renewal, calendar, and utility signals in real time →

The Track of the Day

”Artificial intelligence must enhance human judgment, not replace it. Technology delivers the most value when it strengthens human relationships rather than replacing them.”
, Moody's Summit 2026

Today's set: ”Take Me to the River” by Talking Heads, cued at the moment the headline acts have cleared the booth and the working tracks are carrying the room. The model-lab funding speculation was loud this week. The SaaS-rout reversal, the EU calendar rewrite, the utility consolidation, and the industrial-AI Series B were the tracks that actually moved the floor. The operator who waits for the headline track while the renewal, compliance, energy, and engineering layers keep shipping is going to play to a thinner room by Q3. The one who walks into Thursday with all four rows already on the dashboard is headlining the rest of the cycle.

Yves Mulkers, your data DJ, mixing 190,000 articles into the tracks that actually matter.

We scanned 190,000 articles this week so you don't have to. Data Pains → Business Gains.

Published: May 20, 2026 | Curated by Yves Mulkers @ Ins7ghts