SE2 counts sockets and ignores cores; Enterprise Edition counts every core times a core factor, and that metric flip is where the sticker shock lives. This model shows exactly how your estate reprices, where the traps sit, and when the upgrade is genuinely justified.
SE2 counts sockets and ignores cores; Enterprise Edition counts every core times a core factor, and that metric flip is where the sticker shock lives. This model shows exactly how your estate reprices, where the traps sit, and when the upgrade is genuinely justified.
After 25 years negotiating Oracle Database deals, I can tell you the single fact that decides whether an SE2-to-Enterprise-Edition upgrade is a rounding error or a seven-figure event: SE2 licenses by occupied socket, and Enterprise Edition licenses by physical core times a core factor. Nothing else matters as much. On the April 16, 2026 Technology Global Price List, Enterprise Edition (EE) lists at $47,500 per Processor and $950 per Named User Plus (NUP), with a 25 NUP per Processor floor. Standard Edition 2 (SE2) lists at $17,500 per occupied socket (or $350 per NUP, 10 NUP per server minimum), cores ignored entirely, capped at two sockets per server.
That structure means SE2 is deliberately blind to core count. On a server with two 32-core processors, SE2 requires exactly two Processor licenses, full stop. EE on the identical hardware counts 64 physical cores, applies the 0.5 Intel/AMD core factor, and lands on 32 Processor licenses. Same iron, sixteen times the license count. If you want the full constraint picture behind why estates get forced off SE2 in the first place, start with our Oracle Standard Edition 2 migration decision guide.
SE2 counts two sockets and stops. Enterprise Edition counts sixty-four cores and multiplies. That is not a price increase, it is a different unit of measure.
Here is the repricing engine working on real hardware, all figures at list before any discount. Oracle's rounding rule matters: all fractional license counts round up to the next whole number, so there is no partial-license relief.
| Server (2 sockets) | Physical cores | Core factor | EE Processor licenses | EE list license | EE annual support (22%) |
|---|---|---|---|---|---|
| 2 x 8-core Intel | 16 | 0.5 | 8 | $380,000 | $83,600 |
| 2 x 16-core Intel | 32 | 0.5 | 16 | $760,000 | $167,200 |
| 2 x 32-core Intel/AMD | 64 | 0.5 | 32 | $1,520,000 | $334,400 |
| 2 x 64-core IBM POWER | 128 | 1.0 | 128 | $6,080,000 | $1,337,600 |
| 2 x 128-core AMD EPYC | 256 | 0.5 | 128 | $6,080,000 | $1,337,600 |
Read those last three rows carefully. The IBM POWER factor of 1.0 is double the Intel/AMD 0.5, so a 64-core-per-socket POWER box reprices to 128 licenses, identical to a 128-core-per-socket EPYC machine that carries twice the raw cores. Platform choice alone doubles license count and doubles annual support. That is the platform trap in numbers, and it is why hardware refresh decisions must be run past licensing before the purchase order, not after.
One more point that catches buyers off guard: AMD EPYC 9004/9005 and Intel Xeon 6 are not named individually on the Oracle Processor Core Factor Table. They land on the generic 'Intel and AMD multicore x86' entries at 0.5 by default. Oracle's most recent table refresh (January 28, 2026) added a long list of Xeon SKUs, all at 0.5, so for x86 there is no factor-based relief coming.
The nominal list gap between EE at $47,500 and SE2 at $17,500 is roughly 90.8% per unit, but per-unit comparison is misleading because the units differ (per socket versus per core times factor). More useful is the net reality. In our 2026 negotiations, EE Processor net prices land between $12,000 and $28,000 per processor depending on deal size, competitive tension (a credible PostgreSQL or Microsoft alternative), and preparation. That is a 40% to 75% discount off the $47,500 list.
So the honest reframe is this: the five-year arithmetic, once you apply achievable discounts, support at 22% per year, and the fact that EE consolidates workloads SE2 cannot, turns a headline 90.8% gap into a real gap closer to 50% to 70%. That is still a large number on a big estate, but it is negotiable, and it is the difference between walking into a renewal defeated and walking in with a target. Never accept EE list as the basis for anything.
EE list is theatre. Enterprise buyers in 2026 pay $12,000 to $28,000 net per processor. If your quote starts at $47,500 and stays there, you have not negotiated, you have surrendered.
Every Oracle Database line prices NUP at exactly one-fiftieth of the Processor price. On EE that is $950 against $47,500. The math is unforgiving: the Processor metric wins the moment you exceed 50 real named users per Processor license. Below that, NUP can be dramatically cheaper, but watch the floor: EE enforces a 25 NUP per Processor minimum, so a 32-processor server carries a 800-NUP floor whether or not you have 800 users.
SE2's NUP economics are gentler ($350 per NUP, 10 NUP per server minimum), which is one reason SE2 survives for small, user-bounded departmental systems. Once you cross into EE, model both metrics against actual named-user counts before you commit. On consolidated, high-user, or internet-facing databases, Processor almost always wins and NUP is a trap that inflates on audit.
You do not upgrade to EE for fun. You are pushed off SE2 by one of four hard boundaries. First, the two-socket ceiling: SE2 cannot be licensed on any server (physical or virtual host) with more than two populated sockets. Second, the 16-thread cap: each SE2 instance is limited to 16 CPU threads of user execution, enforced by the software, not merely by contract. Third, RAC desupport: Real Application Clusters was removed from SE2 in Oracle Database 19c, so multi-node active-active clustering forces EE. Fourth, accidental use of EE-only options, covered below. Our deep dive on the SE2 two-socket and 16-thread cap walks through exactly when these bind, and the SE2 RAC removal in 19c analysis covers the high-availability alternatives that avoid an EE jump.
The critical buyer insight: none of these is a reason to accept EE list pricing, and several have escape routes short of full EE. If you are hitting the thread cap, a smaller number of properly sized EE processors, or Data Guard instead of RAC, may cost a fraction of what Oracle's account team proposes. Interrogate whether you need EE across the estate or on one server.
This is the most common and most expensive. Take a departmental system bought on SE2 in year one, on a two-socket, 16-core box, for $35,000 plus $7,700 support. In year three the business doubles and the refresh lands on a four-socket, 64-core chassis. Four occupied sockets puts SE2 out of scope regardless of workload behaviour. The count becomes 32 EE processors: 64 cores at 0.5, so $1,520,000 at list before any discount. The original SE2 spend is stranded. Any credit toward EE is a negotiated trade-up, not a contractual right. Demand a migration credit in writing, and never buy the four-socket box before the EE deal is signed.
When you retire SE2 licenses that share a support set with licenses that remain, Oracle's repricing clause can reprice the surviving licenses to current terms and eat the saving you thought you booked. Check the support set boundaries before you cancel anything. This clause is why 'we'll just drop the old SE2 support' rarely produces the reduction the CFO was promised.
Because SE2 ignores cores, buyers sometimes purchase a 32-core two-socket server intending to run SE2, then find half those cores are unusable by the database thanks to the 16-thread cap. You have paid for silicon Oracle will not let the SE2 instance touch, and if you later flip to EE, every one of those 32 cores becomes a chargeable unit. Right-size the hardware to the license model, not the other way around.
Oracle's DBA_FEATURE_USAGE_STATISTICS view logs any use of EE options or packs on a Standard Edition database. If an audit finds usage, you must upgrade to EE and pay per processor for those features, retroactively. The default posture is permissive: in our estate sweeps, the control parameter that blocks pack access sat at its permissive default in roughly 80% to 90% of databases. This is the quietest, most avoidable path to an EE bill. Lock the parameter down now; our note on enterprise features that sneak into an SE2 estate lists the specific offenders.
EE is a doorway, not a destination. Once on EE, the options layer on at 2026 list: Partitioning $11,500 per processor, Diagnostics Pack $7,500, Tuning Pack $5,000, RAC $23,000, Advanced Compression $11,500, Advanced Security $15,000. On a 32-processor server, Partitioning alone adds $368,000 at list. Price the full stack you will actually enable before signing, and license options only on the processors that use them.
The upgrade earns its cost in a narrow set of cases. It is justified when you are consolidating many SE2 databases onto fewer, larger EE hosts and the license reduction from consolidation (fewer sockets/cores in aggregate, better negotiated discount) outweighs the per-core premium. It is justified when you have a genuine, non-negotiable requirement for active-active RAC, partitioning at scale, or advanced security controls that SE2 simply cannot provide. And it is justified when the workload has outgrown the 16-thread cap and no re-architecture (sharding, read replicas, offloading reporting) is cheaper.
It is a trap in every other case: when the account team proposes EE list, when a hardware refresh (not a workload need) is the trigger, when you could stay on SE2 with a two-socket high-clock-speed box, or when the real answer is to leave Oracle entirely. Before you sign, model the alternatives honestly, including what you actually lose moving SE2 to PostgreSQL and whether an aging SE2 estate belongs on third-party support rather than an EE migration at all. A credible exit is your single strongest lever on the EE net price.
EE is justified when consolidation or a hard technical need drives it. It is a trap when a hardware refresh, an audit finding, or a list-price quote drives it.
The socket-to-core flip is the most expensive four words in Oracle's Database price list. Understand it before Oracle's account team frames the conversation, and you will decide the outcome. Walk in without the model, and the core factor will decide it for you.
On a two-socket, 64-core Intel or AMD server, SE2 lists at $35,000 (two sockets at $17,500). Enterprise Edition on the same box counts 64 cores times a 0.5 core factor, so 32 processor licenses at $47,500, which is $1,520,000 at list. The metric flip from socket to core is what produces that gap, though EE net pricing is typically 40% to 75% below list.
The core factor is a multiplier Oracle applies to physical cores to derive Enterprise Edition processor licenses. Intel and AMD x86 chips are 0.5, IBM POWER is 1.0. On a 64-core server, 0.5 gives 32 licenses on Intel/AMD but 64 on POWER, double the cost. SE2 ignores cores entirely, so the core factor only bites once you move to EE.
Only if you negotiate it. There is no contractual right to trade SE2 spend toward Enterprise Edition; any credit is a discretionary commercial concession. Get it in writing before signing, and be aware that retiring SE2 licenses in a shared support set can trigger a repricing clause that erases the saving.
Four hard boundaries force the move: more than two populated sockets per server, workloads exceeding the 16-thread-per-instance cap, a requirement for RAC (desupported on SE2 since 19c), or discovered use of EE-only options and packs. A hardware refresh that lands on a larger chassis is the most common and most avoidable trigger.
No. In our 2026 negotiations, enterprise buyers pay $12,000 to $28,000 net per processor, a 40% to 75% discount off the $47,500 list. Deal size, credible competitive alternatives, and preparation drive the number. Never accept list as the basis for any calculation.
EE options layer on top per processor at 2026 list: Partitioning $11,500, Diagnostics Pack $7,500, Tuning Pack $5,000, RAC $23,000, Advanced Compression $11,500, Advanced Security $15,000. On a 32-processor server, Partitioning alone adds $368,000 at list. License options only on the processors that actually use them.
Oracle prices Fusion ERP Cloud per employee, not per user, which inflates true cost. The buyer side guide to module economics and the modernization discount.
Gated with a work email on the download page. No sales follow up you did not ask for.
Get the White Paper →
500+ enterprise clients. 11 vendor practices. Industry recognized. One conversation can change what you pay for the next three years.
One buyer side briefing a week. Renewal signals, audit moves, and the levers that work. No vendor spin.
