Aeris, the IoT connectivity company behind the IoT Accelerator platform, has added three eSIM capabilities built around the GSMA's SGP.32 specification and is positioning them for what the industry has started calling physical AI. The announcement on 1 October covers an SGP.32 orchestrator, a single-SKU eSIM for global deployment, and a managed eSIM service that handles remote profile switching on the customer's behalf.
SGP.32 is the GSMA standard that lets a connected device change mobile operator remotely without anyone touching it. The older specifications were built for different situations: SGP.02 for machine-to-machine fleets depends on SMS, and SGP.22 assumes a phone with a screen and a person to tap it. Neither suits a sensor, a tracker or a robot with no interface and a ten-year service life.
Aeris puts 110M devices on its platform, of which 44M use eSIM, a 40% share it contrasts with a market average nearer 12%. It lists 30 operators pre-integrated across all three eSIM generations, more than 7,000 enterprise customers, and devices attaching in over 200 countries.
What The Three Pieces Do
The orchestrator is the substantive part. It manages profile lifecycles through policies and events rather than device by device, which is the difference between an operator change being a project and being a rule that fires when a condition is met.
Single SKU addresses a manufacturing problem rather than a network one. A maker builds one hardware variant, ships it anywhere, and the local operator profile is assigned after the device arrives, which removes the need to forecast demand by market and hold separate inventory for each.
The third piece is a choice about who runs the eSIM IoT Manager, the component that issues profile instructions. Aeris will operate it as a managed service or integrate a third-party one, which matters more than it sounds, because that component is where the control sits once the standard itself is settled.
Why Manufacturers Care About One Variant
Regional SKUs carry costs that do not appear on a connectivity invoice. Each variant needs its own certification paperwork, its own stock, its own forecast and its own line in a bill of materials, and a unit built for one market cannot be diverted to another when demand moves.
Assigning the profile after deployment collapses that into one build. For companies shipping into several regions at once, which is most of the robotics and vehicle telematics market, this tends to be a larger saving than the data tariff.
SGP.32 Is Already A Crowded Field
The press release describes extending leadership, which is a positioning claim rather than a technical one. SGP.32 moved from specification to product across most of the industry earlier this year.
Telenor IoT began commercial deliveries of standardised SGP.32 SIMs in April, emnify launched programmable SGP.32 across its network in March, and Tele2 IoT showed an end-to-end deployment with IDEMIA and Cisco at Mobile World Congress. Kigen has GSMA-certified hardware in several form factors and signed partnerships with KORE in April and Onomondo in July, Eseye folded SGP.32 into its AnyNet platform on 22 April, and Soracom reached general availability on 8 July. The GSMA published version 1.3 of the specification on 28 May, with 1.2 still the baseline for formal certification.
What that leaves is a market where compliance is no longer the differentiator and the orchestration layer is. Aeris is competing on the management software, which is a reasonable place to compete and a different claim from being first.
The Adoption Numbers Are Still Early
Forecasts for SGP.32 are steep and start from a small base. ABI Research counted 2.89M profile downloads in 2025 and projects 194M by 2029, with the sharp climb arriving from 2027. Kaleido Intelligence expects roughly 50M SGP.32-compliant eSIMs under management by 2027.
The installed base today remains overwhelmingly SGP.22 and legacy SGP.02, which is why Aeris emphasises pre-integration across all three. For most buyers the practical question is not whether to move to SGP.32 but how a mixed estate gets managed for the several years it takes to turn over.
Does Physical AI Change The Connectivity Problem
The framing deserves examination. Chief product officer Jon Connet tied the launch to the rise of physical AI, and Orange Business, whose smart mobility executive vice president Elisabeth Reille supplied a supporting quote, called the SGP.32 transition a strategic enabler rather than a technical upgrade.
The provisioning mechanism is the same one asset trackers and smart meters needed. What a robot or an autonomous vehicle changes is the consequence of getting it wrong. A tracker that loses connectivity reports late; a machine acting on what it senses may stop working or keep going on stale instructions, and it is more likely to be crossing borders continuously while nobody can reach it physically.
That raises the value of remote profile management without changing what it is. Physical AI is the demand story here, and SGP.32 is the plumbing that was going to be needed anyway.
What Buyers Should Ask Before Signing
Four questions decide whether this delivers the independence it promises. Which eSIM IoT Manager runs the fleet, and can it be moved to another provider later, since the freedom SGP.32 creates at the operator level can quietly reappear as a dependency one layer up.
Which certification baseline the hardware meets, given the gap between the published 1.3 and the 1.2 used for certification, and which eUICC vendor supplies it, with Thales, G+D, Kigen and IDEMIA all named as compatible here.
What happens to the existing fleet, because most estates will run three eSIM generations at once for years. And what obligations come with long-lived devices under the EU Cyber Resilience Act, whose security assurance requirements apply to products placed on the European market regardless of how the connectivity is provisioned.
The Standard Is Settled, The Control Point Is Not
SGP.32 has done what it was meant to do, which is remove the technical excuse for locking a device to one operator for its lifetime. Every serious platform now supports it, and that part of the argument is over.
The contest that remains is about who runs the management layer and on what terms, which is where Aeris has placed this launch. Buyers evaluating it should read the orchestration claims closely and the physical AI framing lightly, because the first is what they are actually purchasing.
