The perpetual debate is over. The saving is admitting how few people open the software each day.
Autodesk stopped selling new perpetual licences and replaced concurrent network sharing with named user subscriptions, so the choice is no longer perpetual against subscription but which subscription shape fits real demand. The trap is the straight conversion: a network pool where ten engineers shared three licences becomes ten named users, and the estate pays for a peak that never existed on any single day.
Prepared by Redress Compliance · August 10, 2026 · Autodesk advisory. Based on 15 to 25 Autodesk estate reviews, 2024 to 2025.
Executive summary
The straight conversion from network seats to named users overpays, and it overpaid in 6 of 10 estates we reviewed. Under the old model a pool was shared, so ten engineers might run on three licences.
Converting each of those ten to a named user subscription prices a concurrency peak that never occurred, because the shared pool was hiding how few people actually opened the software on any given day.
The named user count should reflect distinct active users measured over time, not the historic seat list and not the vendor's continuity recommendation.
Roughly a third of seats price better on Flex than on a subscription. Flex charges a daily token whenever a user opens a product, which makes it materially cheaper for the occasional tail.
In our file 20 to 35 percent of seats were held as full named user subscriptions where Flex tokens would have cost far less.
The work is measurement rather than negotiation: capture days of use per user per year, find the break even between tokens and a subscription for each usage band, and assign each user to the cheaper model.
Usage segmentation cut annual subscription cost 12 to 22 percent, a median of 18 percent. The cheapest estate almost always mixes both models: heavy daily users on named subscriptions, with a shared Flex token pool covering the long tail of occasional access across the team.
That mix is the entire buyer side win, because the licensing model itself is fixed and not reversible. There is no route back to perpetual, so the leverage sits in sizing rather than in resisting the change.
Uncoordinated renewals raised the effective annual bill 8 to 15 percent through overlapping terms. Co terming seats onto a single renewal date simplifies both the spend and the negotiation, and it removes the overlap that quietly inflates the run rate.
A multi year term is the other timing decision: it locks price against increases but reduces flexibility, so it is worth taking only where the seat count is stable enough that you will not be paying for people who have left the project.
Access options against usage patterns
| Usage pattern | Best option | Why |
|---|---|---|
| Daily full use | Named user subscription | Lowest cost for heavy users |
| Occasional use | Flex tokens | Pay only for the days actually used |
| Mixed team | Subscription plus Flex | Match cost to demand across the tail |
| Legacy perpetual | Plan the conversion | No new perpetual is sold |
The network sharing gap is where the money is, and it is invisible unless somebody measures it. A named user licence follows a person for a term, so the correct count is distinct active users rather than the historic concurrent peak or the number of people who once had access to the pool.
Measure days of use per user per year before setting any count, because that single dataset decides both the subscription number and the Flex population, and it is the only evidence that survives contact with a reseller recommending like for like continuity.
The token mechanics in detail sit in the Flex token pricing guide.
Building the mix that actually costs least
- Measure first. Capture distinct active users and days of use per user per year across the estate, which is the dataset every other decision depends on.
- Find the break even per usage band. There is a threshold where Flex stops being cheaper than a subscription; place frequent users above it and occasional users below it.
- Size named subscriptions to heavy daily users only, and route the rest to a shared Flex token pool rather than defaulting everyone to a seat.
- Co term the renewals onto one date, because overlapping terms added 8 to 15 percent to the effective annual bill in our file for no benefit at all.
- Weigh a multi year term honestly: price protection against the risk of paying for seats the team no longer needs. The enterprise agreement route sits in the Autodesk EBA negotiation guide.
The Autodesk audit defence brief
The compliance posture that sits alongside the subscription decision: entitlement evidence, usage records, and the buyer side response when Autodesk reviews the estate.
Get the white paper →The cost curve, and the timing decisions around it
A perpetual licence was a one time cost plus annual maintenance, while a subscription is recurring with no end, so for a long held seat the cumulative cost crosses above the old model within a few years.
That crossover is the honest framing of the change, and it is also why the seat count matters more than it used to: an oversized perpetual estate wasted capital once, whereas an oversized subscription estate wastes it every year for as long as nobody rechecks the numbers.
Two timing decisions sit on top of the curve.
Co terming aligns every seat onto a single renewal date, which removes the overlapping terms that raised the effective annual bill by 8 to 15 percent across the estates we reviewed and, just as usefully.
Consolidates the negotiation into one event instead of several small ones the vendor handles separately.
The multi year decision trades price protection against flexibility: locking a rate is worth real money in a market where list prices move, but it also locks the seat count, so it suits a stable engineering population and works badly for project based teams whose headcount swings with the workload.
Neither decision changes the licensing model, which is fixed, and that is precisely why they matter: the buyer side win here is sizing and timing, not resistance. The comparison against alternative platforms sits in the Autodesk versus Bentley and Hexagon comparison.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across Autodesk estates, 2024 to 2025
The standard reseller advice is to convert every former perpetual and network seat into a named user subscription for continuity, treating it as a like for like swap. We disagree, because the old network sharing meant far fewer distinct daily users than seats:
Annual subscription cost removed by usage segmentation on mixed usage teams, before any discount conversation with the vendor.
Share of seats held on full subscriptions where token based access would have cost materially less across the year.
Three patterns recurred: full named user subscriptions held by occasional users where Flex would have cost far less on 20 to 35 percent of seats, former network sharing masking real demand so the named user count was set too high at conversion.
And uncoordinated renewal dates creating overlapping terms that raised the effective annual bill 8 to 15 percent.
The buyer side move is to measure distinct active users and days of use first, size named subscriptions to the heavy users, route the rest to Flex, and co term the whole estate onto one renewal date. Autodesk closed the perpetual door, so the old debate is over and the live saving is arithmetic.
The wider library sits in the Autodesk practice.
Your first five moves
- Measure distinct active users and days of use per user across the estate, because the shared network pool hid how few people opened the software on any given day.
- Size named user subscriptions to heavy daily users only, rather than converting the historic seat list one for one as continuity advice recommends.
- Route the occasional tail to a shared Flex token pool, since roughly a third of seats in our file priced better on tokens than on a subscription.
- Find the Flex against subscription break even for each usage band, and assign every user to the cheaper side of it rather than to their old entitlement.
- Co term renewals onto one date and weigh the multi year trade, because overlapping terms added 8 to 15 percent for nothing. The Autodesk practice runs the segmentation and the renewal with you.
Frequently asked questions
Can you still buy perpetual Autodesk licences?
No. Autodesk retired new perpetual sales and moved to a recurring subscription model, so the choice is now between subscription shapes rather than between perpetual and subscription.
Existing perpetual holders face a conversion decision, and the leverage sits in how the converted estate is sized rather than in resisting a model change that is not reversing.
Why does a straight conversion from network licences overpay?
Because the old concurrent model shared a pool. Ten engineers might have run on three licences, so converting all ten to named user subscriptions prices a peak that never occurred on any single day.
In roughly 6 of 10 estates we reviewed, the network sharing meant far fewer distinct daily users than seats, and the straight conversion paid for the difference.
When does Autodesk Flex pay off?
For light and occasional users. Flex charges a daily token whenever a user opens a product, so somebody who works in the software a handful of days a month consumes far less than a full subscription costs.
There is a usage threshold where a subscription becomes cheaper, so map days of use per user per year and place each person on the cheaper side of it.
How much can usage segmentation save?
In our file it cut annual subscription cost by 12 to 22 percent on mixed usage teams, a median of 18 percent, and it required no discount negotiation at all.
The saving comes from two places: named user counts sized to distinct daily users rather than the old seat list, and the occasional tail moved onto tokens instead of full subscriptions.
Does the subscription cost more than perpetual over time?
For a steady, long held seat, yes. A perpetual licence was a one time cost plus annual maintenance while a subscription is recurring with no end, so cumulative cost crosses above the old model within a few years.
That is exactly why seat sizing matters more now: an oversized subscription estate wastes money every year rather than once.
Why does co terming renewals matter?
Uncoordinated renewal dates create overlapping terms, which raised the effective annual bill by 8 to 15 percent across the estates we reviewed for no benefit.
Aligning every seat onto a single renewal date removes the overlap and consolidates the negotiation into one event, instead of several small ones the vendor can handle separately and quietly.
Is a multi year Autodesk term worth signing?
It depends on how stable your population is. A multi year term locks price against increases, which is worth real money when list prices move, but it also locks the seat count.
It suits a stable engineering population and works badly for project based teams whose headcount swings, because you keep paying for people who have left the work.
