Cut engineering costs without killing the roadmap: what to cut first

When your runway shrinks and the board asks for a cut, cut engineering costs in order of how reversible each line is and how little roadmap it touches. In our judgment that means tools first, then cloud, AI seats, contractors, open reqs, and headcount last. Payroll is the biggest line, so the first three rarely close the gap alone. Decide the number you're cutting to, and the roadmap you refuse to give up, before you pick a single line. More leadership guides sit in Startup CTO guides: leadership, diligence and the first hires.

How much do you need to cut, and by when?

No benchmark tells you how much to cut. The inputs are yours: cash in the bank, current net monthly burn, and the number of months of runway the board wants.

The arithmetic is short. Allowed monthly burn is cash divided by target months. The cut is current burn minus allowed burn. With illustrative numbers (not a benchmark): 3.0M in cash and 250k monthly burn is 12 months. To reach 18 months, burn must fall to about 167k, so you need roughly 83k a month out. And since savings take weeks to land, a cut that starts in month 3 has to be bigger than 83k to make the same date.

Then ask the question the CEO will ask anyway, as a hypothetical: what would you drop if we had to cut 30%? Answer it before you're asked. CIO.com's 2021 piece on IT cost-cutting mistakes makes the timing point: waiting for a mandate narrows your options to reactive cuts. In our judgment, the CTO who arrives with a ranked list picks the cuts. The one who waits is likelier to be handed a percentage.

What should you protect before you cut anything?

Write down what stays whatever happens. Three categories qualify.

Roadmap items tied to revenue: the feature a signed customer is waiting for, the integration that unlocks a renewal. Security and compliance work. And commitments you've made to customers in writing, such as SLAs and migration dates.

The same CIO.com article warns that hasty cuts open performance gaps, and that cutting cybersecurity adds risk. Take that as a reason to put security on the protected list, not as a reason to freeze everything. A protected list also gives you a way to say no to the cut that looks cheap and costs a customer.

In what order should you cut engineering spend?

The ranking below is our judgment, not a measured result. It sorts lines by how easily you can undo the cut, how fast the money stops leaving, and how much roadmap you lose.

Line	Reversibility	Speed to save	Roadmap risk	Check before cutting
Tools and SaaS overlap	High, resubscribe	Next billing cycle	Low	Who actually logged in last month?
Cloud waste and commitments	High, unless you commit	Days to weeks	Low to medium	Is anything idle, oversized or unowned?
AI coding seats	High	Next billing cycle	Low, but measure	Does the seat show up in cycle time or review load?
Contractors and agencies	Medium, notice and ramp	Weeks	Medium to high	What ships next quarter because of them?
Open reqs and hiring plan	Medium, restarting costs time	Immediate (no spend yet)	Medium	Which unfilled role blocks a protected item?
Headcount	Low	Months	High	Do you still have the skills to ship the protected list?

Tools and SaaS overlap

Start here because a mistake costs you a resubscription. Export every subscription from the last quarter's card and invoice data and mark duplicates: two CI services, three monitoring tools, an issue tracker nobody opened since the migration. Look at per-seat tools for seats belonging to people who left. Nobody argues about a cancelled licence, so this cut needs no meeting.

Cloud spend

Flexera's 2026 State of the Cloud report estimates that 29% of cloud spend on IaaS and PaaS is wasted. That's a survey estimate across respondents, not a measurement of your account, but it tells you where to look first: idle instances, oversized databases, orphaned storage, staging environments running at night.

The trap is the discount. AWS says Savings Plans offer prices up to 72% below On-Demand, depending on plan type, in exchange for a commitment to a consistent hourly spend over a one- or three-year term. A company trying to shrink should be careful with that trade. A three-year commitment bought at the wrong moment turns a variable cost into a fixed one just when you want flexibility. Right-size first, commit later and shorter, if at all.

AI coding seats

Seats are a small line next to payroll; our guide to what AI coding tools cost per seat has the list prices. Don't cancel by default. In our judgment, a seat that saves a senior engineer a couple of hours a week is worth more than it costs. Check usage, drop the seats nobody opens, and keep the rest.

Contractors and agencies

This is often the cheapest payroll-like line to end, since the cost usually stops after the contract's notice period. It's also the line most likely to hit the roadmap, because contractors are often on the work that can't wait. List what each one is building, who on the team could absorb it, and the date it matters. If a contractor holds a protected item, keep them. Whether contractors count against a freeze is a separate policy question, covered in our guide on whether contractors count against a hiring freeze.

Open reqs and the hiring plan

An unfilled role costs nothing yet, which makes it the gentlest payroll cut. The catch is that you restart the search later, and a good candidate may not wait. Close the roles that don't touch the protected list and pause the rest with a written restart trigger.

Headcount

Last, because it's the least reversible and slowest to pay back (next section). It's also, in our judgment, usually the only line big enough to move the date. ICONIQ's engineering report says personnel is usually about 70-80% of total engineering spend. The report doesn't say which company stage that covers in the section we read, so your split may differ. If tools, cloud and seats add up to a few percent of the bill, you already know where the money is.

What does a layoff actually save, and when?

Less than the payroll line says, and later. Some cost-cutting advice lists layoffs as the fastest, highest-impact option and stops there. Fastest to decide is not fastest to save.

Savings start after notice ends, and you pay severance on the way out. Both depend on the country. Notice periods and severance rules differ across Europe, so the real cost of the layoff route depends on where each person is employed. Ask your employment counsel for the figures per person, and add them to your 30-day model; this isn't legal advice.

For larger cuts there's an EU floor. Council Directive 98/59/EC on collective redundancies says dismissals notified to the competent public authority shall take effect not earlier than 30 days after that notification. Member States can let the authority shorten that period, and fixed-term contracts and public bodies are excluded. It applies only once the cut meets the Directive's thresholds, and national law implements it differently. A small cut at a small company may fall below them, so check with counsel which side of the line you're on.

The practical point: cash saved in the first quarter after a layoff is smaller than the annual salary figure suggests. A cut that has to extend runway by a particular month needs to be decided earlier than the other lines on this list. And it's the one you can't undo. LeadDev's 2026 survey of 600 engineering leaders found 33% reported layoffs and 35% hiring freezes, similar rates to 2025, so you aren't alone. Rehiring the engineer who knew the billing system is still expensive.

Layoffs vs a blended team: how do you keep the roadmap shipping?

There's a third option between "keep everyone" and "cut people": change the mix. Keep an in-house core that owns architecture and the protected items, and cover continuing feature work with contract or nearshore capacity paid by the hour. HighCircl has no minimum hours; check any other vendor's minimum.

Do the arithmetic with your own numbers, not ours. For each role you're considering replacing:

Current monthly cost: fully loaded salary, including employer contributions, equipment and tools, divided by 12.
Replacement monthly cost: hourly rate times the hours you need per month.
Ramp cost: weeks of reduced output plus the time your core team spends onboarding.
Exit cost: notice and severance for the role you replace, which varies by country.

Compare the monthly figures only after you've added ramp and exit costs once. In our judgment, hiring and ramp are real, so this makes sense mainly for work that continues for six months or more. For a three-month stretch, cut scope instead. Our guide on ranking scope by value and cutting to a releasable slice shows the method; here the scope shrinks because the money did, not because the date slipped.

How to cut engineering costs in 30 days

1. Week 1: inventory every line

List payroll, contractors, cloud, tools, AI seats and open reqs with monthly cost, owner and renewal date. Checkpoint: one sheet that adds up to your actual burn within a few percent.

2. Week 2: cut tools and cloud

Cancel duplicated and unused subscriptions, delete idle cloud resources, right-size what's left, and hold off on any long commitment. Checkpoint: a dated list of savings that have already stopped billing.

3. Week 3: rank the roadmap and cut scope

Write the protected list, rank the rest by value, and drop from the bottom. Checkpoint: a roadmap smaller than today's, agreed with product and the CEO.

4. Week 4: decide contractors, open reqs and headcount

Run the remaining gap against the table. If the first three steps close it, stop. If not, decide the contractor, req and headcount cuts with their notice dates and the cost in your model. Checkpoint: a written plan with the number, the date it takes effect, and the work it removes.

5. Tell the whole team once

Announce the plan to everyone at the same time, with the reason and what the roadmap now includes. Piecemeal news can cost you your best people. Checkpoint: one announcement, one Q&A, no second round of cuts planned.

What this means for your runway decision

You're deciding which lines to cut and whether lost capacity gets dropped or replaced with a cheaper mix. In our judgment, take the reversible cuts in the first month, and treat headcount as the last lever, because it saves later than it looks and can't be undone. If payroll is where the gap is, price both routes before choosing: layoff savings after notice and severance, and a blended team after ramp and exit costs.

Judgment, not a rule: don't let the percentage on the board slide set the cut. Set the roadmap you protect, then size the cut around it. If you can't show the board a date at which the cut pays back, you haven't finished the model. For the employer-side numbers behind that model, see our breakdown of the employer-side cost of an in-house vs nearshore developer.

Staffing the roadmap after the cut through HighCircl

If you decide to keep roadmap capacity on a flexible footing, you can hire senior engineers through HighCircl from seven European countries: Poland, Hungary, Slovakia, Serbia, Slovenia, Romania and Spain. Rates run €45-105/hr ($50-115/hr). The 20% margin is capped and added on top of what the engineer earns, so finance can see it. There's no minimum hour commitment, no subscription and no recruitment fee. A deposit of one month's estimated cost is applied to the first invoice. Read the engagement's end terms before you sign.

FAQ

What should a startup cut first to extend runway?

Start with the lines you can reverse: duplicate tools, idle cloud resources, and unused AI seats. Then contractors and open reqs, with headcount last. Protect revenue-linked roadmap items and security work before you cut any of it.

Is it better to cut headcount or cut costs elsewhere?

Cut elsewhere first, because those cuts are reversible and land quickly. But payroll is usually the largest line (ICONIQ's report puts it at roughly 70-80% of engineering spend), so non-people cuts alone often fall short. Price the headcount route properly, including notice and severance, before you decide.

How long until a layoff saves money in Europe?

Longer than the payroll line suggests. Savings begin after notice ends and severance is paid, and both depend on the country. For collective redundancies, Directive 98/59/EC says collective dismissals generally can't take effect earlier than 30 days after notification to the competent public authority (national law may shorten or extend this). Ask employment counsel for per-person figures.

Should you cancel AI coding tool seats to save money?

Not by default. Seats are a small line next to payroll. Check usage, remove seats nobody opens, and keep the ones that cut review or cycle time. Cancelling a useful seat can cost more than it saves.