The SaaSpocalypse of 2026: Why $1 Trillion Vanished from Enterprise Software — and What Every CTO Must Now Decide About Build vs. Buy

The 48 Hours That Repriced an Industry

On February 3, 2026, Jeffrey Favuzza, an equity trader at Jefferies, coined a term that would dominate enterprise software conversations for the rest of the year. He called it the SaaSpocalypse. Specifically, he described the trading as ‘very much ‘get me out’ style — people are just selling everything and don’t care about the price.’ In a single session, roughly $285 billion in market capitalization evaporated from enterprise software stocks. Furthermore, within six weeks the damage approached $1 trillion.

Notably, the S&P 500 Software & Services Index fell nearly 13% in a single day. That was the worst single-day performance on record.

Why this trigger, why now

The trigger was not earnings. Instead, it was a product release. On January 30, 2026, Anthropic quietly pushed eleven plugins to GitHub under the product name Claude Cowork. Specifically, the plugins targeted enterprise workflows directly. Examples include legal research, compliance review, contract drafting, sales preparation, and meeting summarisation. Furthermore, Anthropic had already shipped Claude Code as an autonomous coding agent. Together, the two products made the case visible. Specifically, AI agents could now do what enterprise SaaS does — and they could write the code to replace it.

Importantly, the thesis was not new in February. Microsoft CEO Satya Nadella had argued the same point on a December 2024 podcast. Specifically, he suggested SaaS applications would collapse once AI agents took over the business logic layer. However, what changed in early 2026 was that the thesis stopped being speculation. Notably, autonomous coding agents had reached production quality. Suddenly, a team of two engineers with the right tools could ship what previously required twenty.

The named casualties

The companies hit hardest were not obscure ones. Specifically, HubSpot fell roughly 51% from peak to trough — from approximately $880 per share to around $233. Monday.com declined approximately 44%. Similarly, ServiceNow dropped approximately 36%. Atlassian fell 27% in eighteen trading days.

Furthermore, the analyst downgrades followed. On February 23, 2026, Jefferies analyst Brent Thill issued a sweeping reset report. Notably, he downgraded Workday and DocuSign from Buy to Hold. Specifically, he slashed Workday’s price target from $325 to $150. Across application software broadly, names plummeted 30–55% from year-start through late February.

Importantly, the carnage was concentrated, deliberate, and category-specific. Notably, the market was not panicking indiscriminately. Instead, it was pricing in a specific thesis. Specifically: if AI agents can do what enterprise software does, then enterprise software — as we have known it — is structurally contested.

This post is the operating brief for engineering and procurement leaders making 2026–2027 software decisions. Accordingly, it synthesizes the post-crash data into one diagnosis. The SaaSpocalypse is not the death of enterprise software. Instead, it is the repricing of a specific category of enterprise software. Furthermore, the difference between which vendors survive and which get canceled in the 2027 renewal cycle is now answerable with a matrix.

What Actually Triggered the Crash

The market did not panic over one event. Instead, it panicked because three things converged at once. Notably, the convergence made a previously theoretical risk suddenly operational.

Trigger One: Autonomous coding agents reached production quality

Between late 2025 and early February 2026, several releases crossed a quality threshold. Specifically, Anthropic shipped Claude Code as the first autonomous coding agent that could navigate large codebases. Furthermore, the same company released Claude Cowork as a desktop tool for automating file and task management through natural language. Likewise, OpenAI shipped agent capabilities at similar quality. The combined message was unmistakable. Notably, software development was no longer a labor-bound activity.

Trigger Two: The build cost curve collapsed

The economics that made buying SaaS the rational choice inverted within twelve months. Specifically, the cost to build a useful internal tool dropped from $50,000–$500,000 down to $500–$20,000. Notably, that is a 25-to-100-fold reduction depending on project size. Furthermore, the timeline collapsed in parallel. What used to take six to eighteen months now takes days to weeks. Consequently, when build is faster and cheaper than buying, the calculation that justified per-seat SaaS for a decade no longer holds.

Trigger Three: The builders already knew

The most important detail in the Retool 2026 Build vs. Buy Shift Report was about timing. Specifically, by the time the report dropped on February 18, builders had been quietly replacing SaaS for at least eighteen months. Notably, across 817 enterprise respondents, 35% had already replaced at least one SaaS tool with a custom build. Furthermore, 78% planned to build more in 2026.

Importantly, the Retool data revealed something the market then ratified. Specifically, the categories leading replacement — workflow automation and internal admin tools — were exactly the categories where SaaS vendors had built their per-seat pricing moats. As David Hsu, Retool’s CEO, put it: ‘The markets are finally catching up to something builders have always known: that enterprise AppGen has become a threat to traditional SaaS.’

Why the February timing mattered

The SaaSpocalypse was not a prediction. Instead, it was a market reaction to facts that had been accumulating since mid-2024. Specifically, what made February the trigger was not a new technological capability. Rather, it was the moment institutional investors recognized that the replacement behavior was already underway inside the customer base. Notably, the market stopped pricing SaaS revenue as recurring. Instead, it started pricing it as contestable. Importantly, that repricing happens once — and it happened in 48 hours.

The Numbers Behind the Repricing

Here is the consolidated 2026 picture across market data, build economics, and AI adoption metrics:

Metric	2026 Value	Source
Enterprise software market cap erased (single trading day, Feb 3, 2026)	~$285 billion	Multiple sources; Jefferies framing
Aggregate market cap erased (6 weeks following)	~$1 trillion	S&P 500 software index
S&P 500 Software & Services Index — Feb 3 single-day drop	~13%	Worst single-day on record
HubSpot (HUBS) — peak-to-trough decline	~51%	From ~$880 to ~$233/share
Monday.com (MNDY) — peak-to-trough decline	~44%	Public trading data
ServiceNow (NOW) — peak-to-trough decline	~36%	Public trading data
Atlassian (TEAM) — 18-trading-day decline	~27%	Public trading data
Workday (WDAY) — Jefferies price target cut	$325 → $150	Brent Thill, Jefferies (Feb 23, 2026)
Custom build cost — pre-AI baseline	$50K – $500K	Pre-AI tooling era
Custom build cost — 2026 with AI coding agents	$500 – $20K	AI-augmented engineering
Enterprise teams that replaced ≥1 SaaS tool with custom	35%	Retool 2026 Build vs Buy Shift Report
Teams planning to build more custom tools in 2026	78%	Retool 2026 (n=817)
Gartner global enterprise software spend, 2026	~$1.43 trillion (+14.7% YoY)	Gartner forecast 2026
Hyperscaler AI infrastructure spend, 2026	$660 – $690 billion	Industry estimates

Notes on sources: Atlassian, HubSpot, Monday.com, ServiceNow, and Workday figures verified against public trading data and the Jefferies February 23, 2026 downgrade report (Brent Thill). The $285B figure is widely cited and attributable to Jeffrey Favuzza’s framing on February 3, 2026. The $1 trillion aggregate is the S&P 500 software index measure through mid-February. Retool figures are from the company’s 2026 Build vs. Buy Shift Report (n=817), released February 18, 2026.

Two patterns worth reading carefully

First, the asymmetry between market reaction and underlying customer behavior. Specifically, the 30–55% sector drop is a binary, sentiment-driven event. Conversely, the 35% customer-replacement rate is a slow-cooked operational shift. Notably, that shift took eighteen months to build. Importantly, markets respond at the moment they recognize the pattern. Meanwhile, CFOs were already executing the pattern long before.

Second, the speed of the build cost collapse. Specifically, a 25-to-100-fold reduction in the build cost over roughly twenty-four months. Notably, no technology curve in modern enterprise software has moved that fast. Furthermore, no SaaS pricing model assumes a build cost curve moving at that velocity.

The Counter-Narrative: Why the Panic Is Partly Overstated

The market reaction was extreme. Furthermore, the data is real. Likewise, the build-vs-buy inversion is genuine. However, none of that means enterprise software is dead. Specifically, the strongest counter-arguments from analysts and operators are worth taking seriously. Importantly, they shape which vendors actually survive the repricing and which do not.

Counter-Argument One: Enterprise software spend is accelerating

Jason Lemkin of SaaStr has argued the inconvenient detail buried under the panic. Specifically, enterprise software spend in 2026 is accelerating in absolute terms. Notably, Gartner’s worldwide software spend projection for 2026 sits at approximately $1.43 trillion. Furthermore, that figure represents +14.7% year-over-year growth, with GenAI as the primary driver. Importantly, the aggregate market is growing even as per-vendor revenue is being repriced. Specifically, spend is moving — from horizontal per-seat tools toward systems of record, infrastructure, and AI-native platforms. However, the spend is not shrinking. Consequently, the narrative that AI is killing enterprise software conflates two different things. AI is killing a specific pricing model and a specific category of horizontal SaaS. Meanwhile, it is accelerating spend toward different parts of the stack.

Notably, Scott Galloway publicly called the selloff ‘farcical.’ Furthermore, he announced he was buying SaaS stocks during the rout. Specifically, he argued that companies like Salesforce and Adobe were trading at decade-low free cash flow multiples. Likewise, JPMorgan and Goldman Sachs published research arguing the selloff was overdone.

Counter-Argument Two: Systems of record still have deep moats

ERP systems, identity platforms, payment rails, data warehouses, cloud infrastructure, and security platforms all share a structural characteristic. Specifically, they have deep, multi-year integration into the systems they touch. Notably, that integration is not present in horizontal workflow tools.

For example, replacing an ERP requires migrating a decade of financial transactions, audit trails, regulatory filings, and downstream integrations. Likewise, replacing identity infrastructure means re-onboarding every user across every application. Similarly, replacing the payments rail means renegotiating with banks and processors. Importantly, the moat is not the software. Instead, the moat is what the software is connected to. Specifically, AI agents can replicate the user-facing logic of an ERP. However, they cannot painlessly migrate twenty years of bookkeeping out of one.

Counter-Argument Three: Replacement is slower than repricing

Markets reprice instantly. Conversely, enterprises do not migrate instantly. Specifically, the Retool data shows 35% of enterprises have replaced one SaaS tool — singular. Furthermore, the same data shows 78% intend to build more. However, ‘intend to’ is not ‘have completed.’ Notably, the actual replacement of an enterprise SaaS portfolio takes years, not quarters. Importantly, the market’s 30–55% drop in application software valuations assumes a replacement velocity that the operational reality is not yet matching. Consequently, some of the repriced vendors will recover when actual cancellation rates prove slower than the panic implied. Meanwhile, others will not recover. Specifically, the panic correctly identified that their category was structurally replaceable.

Counter-Argument Four: Even AI labs are SaaS customers

Notably, the most-cited operational rebuttal to the SaaSpocalypse narrative comes from inside the AI labs themselves. Specifically, both OpenAI and Anthropic CEOs have publicly confirmed their organizations use Slack. Furthermore, the companies building the most capable AI agents are themselves enterprise SaaS customers. Importantly, this matters because it demonstrates a ceiling on the disruption thesis. Specifically, if the entities building the technology that supposedly kills SaaS are themselves SaaS customers, then SaaS is not categorically dying. Instead, certain categories of horizontal per-seat SaaS are being repriced. Conversely, system-of-record and infrastructure SaaS remains structurally protected.

The Vendor Casualty List

Across the post-crash analyses, a consistent pattern emerges. Specifically, the market is making category-specific bets, not blanket pessimism. Furthermore, the vendors hit hardest share a common shape. Notably, their per-seat pricing depends on humans executing workflows that AI agents can now execute. Here is the consolidated view:

Vendor	Hit	Why the Market Repriced
HubSpot (HUBS)	~−51% peak-to-trough	Marketing automation logic is replicable; AI agents author and execute campaigns at fractional cost. Most-exposed of the named casualties.
Monday.com (MNDY)	~−44%	Workflow and project management UI is a textbook AI-agent automation target.
ServiceNow (NOW)	~−36%	ITSM workflows are exactly what agentic AI executes natively; workflow-engine layer is the most exposed of the systems-of-record cohort.
Atlassian (TEAM)	~−27% (18 days)	Jira ticket workflows are an AI agent’s ideal automation target; pricing model assumes humans clicking buttons.
Workday (WDAY)	Price target cut $325→$150	Jefferies downgrade Feb 23, 2026 (Brent Thill). HR workflows (PTO, expense, onboarding) are textbook AI-agent automation.
DocuSign (DOCU)	Downgraded to Hold	Jefferies downgrade Feb 23, 2026. Seat-based e-signature model exposed.
IGV ETF (basket)	Application software −30–55% YTD	iShares Expanded Tech-Software ETF reflecting the aggregate application-software repricing.

Two observations for buyer-side strategy

First, the vendors hit hardest sell software to humans who click buttons. Specifically, Atlassian sells Jira to humans who file and update tickets. Likewise, ServiceNow sells ITSM to humans who route support requests. Similarly, Salesforce sells CRM to humans who log activity. Notably, the structural commonality is not the category of software. Instead, it is the assumption that a human will be the one clicking buttons. Importantly, AI agents can now do all of that clicking at fractional cost.

Second, the vendors not on this list share the opposite shape. Examples include Snowflake, AWS, Stripe, and Okta. Specifically, their value is in the data and integrations they enable. Notably, it is not in human workflow execution. Furthermore, the repricing is making this distinction visible at the market level for the first time.

The Vendor Vulnerability Matrix: Which Categories Survive

The most useful tool for any CTO or procurement leader navigating the 2026 renewal cycle is a two-dimensional matrix. Specifically, it maps AI replaceability against switching cost. Notably, every SaaS line item in the budget falls into one of four quadrants. Furthermore, the strategic move differs by quadrant:

Quadrant	Characteristics	Examples	Strategic Move
Safe — Keep Buying	Low replaceability + high switching cost	ERP · Identity · Payments · Data warehouse · Cloud infra · Security	Renew. Negotiate. Optimize.
Contested — Renegotiate	High replaceability + high switching cost	CRM · ITSM · HRIS · Marketing automation · DXP · Productivity suites	Negotiate hard. Reduce seats. Pilot replacements.
Shop Around	Low replaceability + low switching cost	Scheduling · Light DevOps · Transactional email · Survey tools	Comparison shop. Switch on price.
Danger Zone — Build	High replaceability + low switching cost	Internal admin tools · Workflow automation · Form builders · Light iPaaS	Build it yourself. Cancel renewal.

How to use the matrix

The matrix is the single most useful conversation an engineering organization can have about its 2027 budget. Specifically, walk down the SaaS line items. Then, assign each one to a quadrant. Notably, the action plan writes itself.

First, Safe-quadrant vendors get renewed with negotiated discounts. Second, Contested-quadrant vendors get a hard renegotiation, with seat-count reductions and pilot replacement projects to test what could be built internally. Third, Shop-around vendors get switched on price, since the switching cost is low. Fourth, Danger-zone vendors get canceled at the next renewal and replaced with a custom build. Importantly, for most large enterprises, the Danger Zone alone represents 15 to 30 percent of the current SaaS budget. Furthermore, the Retool data suggests that 35% has already started moving.

What the Build Decision Actually Costs (The Other Side)

The build-vs-buy argument is incomplete if it only counts engineering hours. Notably, three operational costs deserve explicit accounting. Importantly, ignoring them is how custom-build initiatives fail in year two.

The compliance and security burden

SaaS vendors handle SOC 2, GDPR, HIPAA, SOX, security patching, audit logging, penetration testing, and disaster recovery. Conversely, custom builds inherit all of that. Specifically, the $500–$20,000 build cost is the engineering cost. Notably, it does not include the ongoing compliance burden. Furthermore, for regulated industries, the annual compliance overhead on a custom-built tool can range from $10,000 to $100,000+ per tool per year. Consequently, that overhead can dwarf the original engineering cost over five years. Importantly, the build-vs-buy decision must include compliance scope. For example, an internal admin tool with no customer data is one calculation. Conversely, a customer-facing workflow tool subject to SOC 2 audit is a substantially different one.

The contract mechanics of SaaS cancellation

Enterprise SaaS contracts have multi-year terms, auto-renewal clauses, and termination penalties. Specifically, ‘cancel renewal’ is not always an option in the current contract year. Furthermore, most enterprise SaaS contracts auto-renew unless terminated 60–90 days before the renewal date. Notably, the operational reality of the Danger Zone move is more layered. Specifically, audit contract renewal dates first. Then, file termination notice within the contractual window. Furthermore, plan replacement before the lapse date. Importantly, a custom replacement that is six weeks behind schedule is a six-week SaaS bill the enterprise did not budget for.

Maintenance and ongoing engineering

Custom builds need someone to maintain them. Specifically, the API changes when an upstream vendor changes their schema. Furthermore, the security patches do not apply themselves. Notably, the feature requests from internal users accumulate. Importantly, the typical pattern is that maintenance cost runs at 15–25% of the original build cost per year. Consequently, a $20,000 build implies a $3,000–$5,000 annual maintenance line indefinitely. For most Danger Zone replacements, this is still substantially cheaper than the SaaS subscription. However, the budgeting must include it. Otherwise, the build looks like a one-time savings rather than the recurring savings it actually is.

The honest TCO comparison

Specifically, the right comparison is not ‘$200K/year SaaS versus $20K build cost.’ Instead, it is ‘$200K/year SaaS versus $20K build + $5K/year maintenance + $20K/year compliance.’ Importantly, even with the full cost stack, Danger Zone replacements typically still cut TCO by 60–80% over five years. However, the budgeting honesty determines whether the savings are real or theatrical. Specifically, projects that budget only the engineering cost fail in year two. Conversely, projects that budget the full stack succeed.

What This Means for Custom Software Development

The build-vs-buy inversion is not just a SaaS market story. Instead, it is a structural opportunity for custom software development. Specifically, an opportunity that has not existed at this scale in twenty years.

Notably, the economics that justified buying SaaS over building custom — high engineering cost, long timelines, scarce talent — have collapsed. Meanwhile, the economics that justified custom over SaaS — competitive differentiation, exact-fit requirements, integration control — remain intact. Consequently, the result is the most favorable build-side economics enterprises have seen since the original web era.

Three concrete shifts in the conversation

For PracticalLogix and the enterprise customers we work with, the 2026 conversation has shifted in three concrete ways.

First, the question is no longer whether to build custom. Specifically, for at least 15 to 30 percent of the SaaS portfolio in any given enterprise, custom is the obvious answer once the matrix is applied.

Second, the question of how to build has reframed itself. Notably, not from scratch, not as a multi-year project. Instead, as an AI-augmented engineering effort that delivers a working replacement in weeks rather than quarters.

Third, the question of who builds is now answerable in three ways. First, pure in-house. Second, pure agency-built. Third, the increasingly common hybrid that pairs in-house product knowledge with external engineering velocity. Importantly, all three are now economically viable. Conversely, pre-2026 they often were not.

The strategic rule for the 2027 budget cycle

Specifically, build what is high-replaceability and low-switching-cost. Furthermore, keep buying what is deeply embedded and hard to migrate. Renegotiate everything in between. Notably, the single biggest mistake CTOs are making in 2026 is treating the entire SaaS portfolio as one decision. Instead, it is twenty to fifty decisions, and each line item has its own quadrant. Importantly, the renewal conversation that goes line by line and applies the matrix captures the savings. Conversely, the conversation that produces a generic ‘reduce SaaS spend by X percent’ target captures none of them.

Practical Takeaways: What to Do This Quarter

For engineering leaders, CTOs, and procurement directors making 2026–2027 software decisions, here is the prioritized action list. Notably, none of these require completing the SaaS-to-custom migration this quarter. However, all of them require starting the diagnostic this quarter.

Diagnostic: audit, identify, pilot

• First, run a SaaS portfolio audit using the matrix.Specifically, list every SaaS line item. Then, plot each on the replaceability-versus-switching-cost matrix. Furthermore, produce a quadrant assignment for each. Notably, this is the single most valuable two-week exercise in engineering this year. The output is a vendor-specific action plan, not a generic cost-reduction target.
• Second, identify your Danger Zone vendors first.Notably, the 15 to 30 percent of the portfolio in the Danger Zone is where the immediate savings are. Examples include internal admin tools, workflow automation (Zapier-class), form builders, and light integration platforms. Specifically, these are the consistent top targets across the Retool data.
• Third, pilot a custom replacement for one Danger Zone vendor.Specifically, pick the single SaaS line item with the highest cost-to-replaceability ratio. Then, build the replacement using AI-augmented engineering. Notably, the operational learnings from one replacement inform the rest of the migration plan. Furthermore, those learnings include timeline, maintenance burden, integration friction, and compliance scope.

Execution: renegotiate and lock in

• Fourth, renegotiate Contested Zone vendors hard.Notably, Salesforce, Workday, ServiceNow, Atlassian, HubSpot, and Marketo are all in active price negotiation with their installed base. Specifically, going into the 2027 renewal with the matrix analysis in hand, the seat-count audit completed, and a credible build-side alternative identified produces dramatically better terms. Conversely, walking into a standard renewal conversation does not.
• Fifth, lock in your Safe Zone vendors with multi-year discounts.Specifically, ERP, identity, payments, data warehouse, and cloud infrastructure are not on the casualty list. Importantly, they are also not getting cheaper. Therefore, a two-to-three-year commitment at this year’s pricing locks in cost before the next round of inflation.

Sequencing: benchmark, replace, communicate

• Sixth, audit your engineering team’s AI-augmented capacity.Specifically, the build-vs-buy economics depend on how fast your team ships with AI tools. Notably, run a benchmark on one realistic project. First, measure baseline hours. Then, measure hours with Claude Code, Cursor, or Copilot at full deployment. Importantly, the ratio is the input to every subsequent build-vs-buy decision.
• Seventh, plan the second-wave replacements.Notably, the first replacement is the proof point. Conversely, the second through tenth are where the real budget savings happen. Specifically, build a multi-quarter sequence. Furthermore, decide which Danger Zone vendor gets replaced when, by which team, with what level of AI augmentation. Importantly, this sequence is the difference between a one-time savings story and a structural cost reduction.
• Eighth, communicate the strategy to the board with the matrix, not the headlines.Specifically, the SaaSpocalypse headlines will lead the board to ask whether you are cutting SaaS spend. Conversely, the right answer is more nuanced. Notably, you are repricing the portfolio, keeping the deeply embedded systems, canceling the structurally replaceable ones, and reallocating the savings to custom builds that create competitive differentiation. Importantly, that story sounds like strategy. Conversely, the headline-cut story sounds like a budget panic.

What This Means for 2026–2027 Budget Decisions

The right framing for the 2026–2027 software budget conversation is not whether to cut SaaS spend. Instead, the matrix produces a more useful question. Specifically: which vendor categories should be renewed, which should be renegotiated, which should be canceled, and what should the cancellation savings be reinvested in?

For most large enterprises, the answer follows a consistent pattern. Specifically, 60 to 70 percent of the current SaaS portfolio stays. Furthermore, 20 to 30 percent gets aggressive renegotiation. Finally, 10 to 20 percent gets replaced by custom builds. Notably, these ranges reflect PracticalLogix’s observed split across enterprise engagements through Q1 2026. They are working ranges, not analyst projections. Importantly, the aggregate savings, redirected into custom engineering, typically produces a net cost reduction of 15 to 25 percent against the prior year’s software line.

How PracticalLogix frames it

For PracticalLogix and the enterprise customers we work with, the framing is this. Specifically, the SaaSpocalypse is not a market panic to be ignored. Instead, it is a structural repricing that creates the most favorable build-side economics enterprises have seen in twenty years. Notably, the enterprises that capture the upside treat the next renewal cycle as a portfolio rebalancing exercise. Conversely, they do not treat it as a horizontal cost-cutting exercise. Importantly, the matrix is the input. The vendor-specific action plan is the output. The 2027 budget is the test.

Conclusion: From Buy Default to Strategic Portfolio

The SaaSpocalypse of 2026 was not the death of enterprise software. Instead, it was the death of a specific assumption that had governed enterprise procurement for fifteen years. Specifically, the assumption that buying SaaS was always cheaper, faster, and lower-risk than building custom. Notably, that assumption was correct for almost the entire SaaS era. However, it stopped being correct in 2026. Furthermore, the market priced the change in 48 hours.

Two shifts now in motion

First, the vendors who depend on the old assumption are losing valuation. Conversely, the enterprises that recognize the new assumption are reallocating their software budgets. Importantly, both shifts are now in motion. Notably, neither is reversing.

Second, participation is no longer optional. Specifically, the economics make participation inevitable. However, the question is whether to participate strategically or reactively. For example, the strategic approach uses a matrix, a vendor-specific action plan, and a multi-quarter migration sequence. Conversely, the reactive approach uses headline-driven budget cuts. Notably, the first approach reallocates spend toward competitive differentiation. The second approach saves money on subscriptions and gives up the chance to use the savings for anything that matters.

The one-time window

Notably, the SaaSpocalypse is a one-time repricing opportunity. Specifically, the next eighteen months are when the portfolio rebalancing happens. Importantly, the enterprises that complete the diagnostic now will reach 2028 with a fundamentally restructured software cost base. Furthermore, they will have an in-house engineering capability that did not exist in 2024. Conversely, the enterprises that wait will face the same migration in 2028 under crisis pressure. Specifically, they will face it with vendors that have hardened their pricing positions and engineering markets that have tightened. Which one your organization is depends on what you do this quarter.