Quite a few things Itanium-related have been cancelled over the years. This is a partial list of them.
Contrary to popular belief, Monterey actually did ship - briefly - as AIX 5L for Itanium. It didn't last, and is included here as a result.
Novell's Modesto operating system would have been an unholy fusion between Netware and a hypervisor OS. Kind of cool, really, and showed obvious inspiration from IBM VM. It does not seem to have ever shipped, and it's unclear how far in development it got.
Sun did the port, then declined to release it. May have been briefly resurrected, then killed again, in the 2004 timeframe.
Tru64/IPF was a casualty of the Compaq/HP merger; HP was going to go forward with only one IPF UNIX, and Tru64 had the smaller customer base.
Also a casualty of the Compaq/HP merger. MPE had a Merced port announced, though it's not clear how far along the port got before being canned. VMS fills roughly the same niche in the product line, but apparently had the less expendable customer base.
An IPF version of SGI's UV supercomputer, the successor to the Altix 400/4000, was planned; it was alluded to in the infamous "Read My FLOPS" blog post from SGI in 2009, and appeared on some customer roadmaps. It never shipped, presumably due to Tukwila's uncompetitiveness.
HP's initiative to give Itanium a soft landing. Kinetic would have involved Kittson being socket-compatible with Xeon, a Xeon port of HP-UX, and mixed Itanium and Xeon cells within a single Superdome chassis.
Sound awesome? I agree. HP feeds off the pain of its users, though, and Kinetic is long dead. See also: Kittson.
A 22nm chip that would have been the last Itanium. Intended to share a socket and a platform (presumably Brickland) with future Xeons. Canned in January 2013, shortly after announcement.
A new Poulson stepping was announced as being "Kittson" in 2017, bringing a clock boost from 2.53GHz to 2.66GHz and no new features. Whether this is a cynical marketing ploy so Intel can say "we delivered Kittson as per the roadmap" or a genuinely new Itanium generation is left as an exercise to the reader.
An HP-UX port to x86 achieved first boot on 23 Dec 2009. It didn't survive. See also: Kinetic.
See Kinetic.
A commodity chipset for the Montecito generation, replacing the elderly E8870. Deader than disco. Intel's own SDVs for that generation used a Hitachi Coldfusion-3 chipset.
Montecito's low-cost counterpart, the successor to Fanwood. Dead. While it likely wouldn't have been technically impressive, Millington's non-appearance arguably marks where Itanium abandoned trying to be a commodity CPU family, especially in conjunction with Bayshore's cancellation.
Millington's successor, a low-cost Tukwila variant.
Tanglewood, before it became Tukwila (an underwhelming I2 rev with an early, power-hungry QPI controller) was defined as a manycore (by the standards of the time) Itanium implementation with a new microarchitecture, built by the ex-Compaq design group. Canned in 2005. I have some suspicion that Tanglewood microarchitectural DNA ended up in Poulson (also a product of the Hudson design group.)
Huawei was announced as an Itanium OEM in 2011, and listed as an OEM during the
Poulson launch presentation in 2012 (but with no hardware.) The emails released
as part of the HP v Oracle lawsuit indicate that Huawei had designed a glueless
8-socket system internally and had a 32-socket system with a custom chipset
under development in Spring 2011, and were interested in licensing HP-UX but
concerned about the future of both HP-UX and Itanium itself. The project's fate
post-2012 is unclear, but "dead" seems to be a safe bet.
Likely a dual-CPU workstation built on the C8000's board design. Canceled when HP withdrew from workstations.
Montecito-capable 1U system canceled shortly before release. Shows up in OpenVMS docs. Probably uses a zx2 chipset based on its codename (Onyx - compare to rx3600 Ruby and rx6600 Sapphire) but may be zx1.