It was called "Mont Blanc" and it was in development with a partner company called STMicroelectronics (no space). ST is a giant Dutch company, and they actually manufacture the Amiibo NFC stack for the Switch.
Mont Blanc went through a couple designs, and the final version before it was scrapped was a quad core ARM 53 CPU cluster combined with a custom "Decaf" GPU. "Decaf" was, essentially, the Wii U's "Latte" GPU stripped of all its Wii/GameCube backwards compatible circuitry, and then cut in half. Mont Blanc's performance target was actually not only lower than the Switch, but lower than the Wii U. Lemme give you some context.
Iwata wanted to unify handheld and TV console
software development. It was too expensive. Mobile GPUs were in their infancy, and the 3DS had a totally different GPU from the DS. Basically Nintendo had to start over from scratch with each new mobile console. But on the TV side, they had been developing their GPU platform all the way back to the N64. Nintendo had top tier 3D software engineers squeezing all this power out of a design they knew like the back of their hand, and
none of that knowledge - or engine work - could be reused on the (more successful) handheld side.
Nintendo launched Project Indy. Project Indy had two goals. One, standardize the handheld GPU platform with the TV platform, or as Iwata put it
"absorb the Wii U architecture". The second was to build a new concept for the Next Handheld, the successor to the 3DS.
On the concept side, Nintendo tried a bunch of Wild Nintendo Shit. One of the feature they played with involved casting from the handheld to the TV...
On the chip side, they began working on "Decaf". Decaf took the Wii U's "Latte" GPU, stripped out all the legacy support, and made it much smaller. The idea being that the Wii U would run Latte, the handheld would run Decaf, and the Wii U's successor would run a big version of Decaf (Half Caf?). Indy would combine Decaf with ARM CPUs, because those were/are the standard for low power, high performance mobile devices. That combined chip was Mont Blanc, which would also include things like memory and IO controllers and all the other bells and whistles that a custom SOC could deliver.
While Nintendo worked on Mont Blanc with ST, the concept for Indy was evolving. "Casting" had latency problems, but "docking" was promising. Somewhere in 2014, docking wasn't just a feature, it became the core concept, and the device was given a name - The Switch. The gigaleak has a document for what a Mont Blanc based Switch would look like, dated pretty late in 2014, in fact.
But at the point at which The Switch wasn't just a 3DS successor but also a Wii U successor, Nintendo actually didn't
need Decaf. The platforms would be unified not by a shared architecture, but by
literally being the same device. If they dropped Decaf, they'd lose their historical investment, but on the other hand, they had designed Decaf to be a mini Wii U in performance, and now it needed to
beat the Wii U to be its successor.
Nintendo was already familiar with the Tegra line, and in 2014, the security team at Nintendo did an analysis of both the Mont Blanc design and the (unreleased) Tegra X1 design for the Switch. TX1 actually had some problems in Nintendo's view, and they were concerned about possible security holes leading to piracy. Nvidia did some fairly last minute overhauls of the Tegra Security Module to pass Nintendo's security review - apparently not even being sure they'd get the contract.*
While it's not documented in the Gigaleak
why Nintendo went with Tegra X1 over Mont Blanc I think the answers are pretty obvious. Once Indy became the successor to not just the 3DS but the Wii U, Mont Blanc couldn't cut it. It wasn't fast enough, but more important, it wasn't
ready. Tegra X1s were already in production at the end of 2014, but Mont Blanc wasn't even ready to tape out.
Nvidia was also
hungry for the contract. Not only did Nvidia redesign the TSM for Nintendo before the contract was signed, they built
NVN before the contract was signed. Nintendo had 15+ years of software development invested in the Wii U's GPU they had to toss, but Nvidia was willing to put their extensive driver experience on their own hardware up for Nintendo's use and put their money where they mouth was.
There are documents from
December of 2014 still talking about a Mont Blanc based Switch, but I suspect in honesty that the deal with Nvidia was de facto done at that point, even if on paper ST was still in the game. 3 months later, in March, Nvidia and Nintendo had not only signed the contract, they had built a functional Switch prototype. Nintendo shifted every Wii U game that wasn't announced over to the Switch, and set a release target of October 2016. They were
booking it.
At some point after the Switch's release, some folks in the fan community put what they new about Project Indy together and started to think of it as a separate console built on Mont Blanc. But Indy
is the Switch, they were always the same project, and Mont Blanc is mostly interesting not for what it could have been but for what it never was.
We have these rumors of a cancelled Pro, and 11 dev kits and spice. Mont Blanc
had devkits - but they were Wii U's wired up to off the shelf ARM chips. And again, these were devkits for Indy,
a device we got, but in a vastly different internal design.
And when imagining a cancelled T239, or a T239 repurposed by Nvidia, consider how ST didn't try to salvage 2 years of development work on Mont Blanc by repurposing a product that essentially had no customers except Nintendo or Nintendo's competitors.
* Side note, this is quite hilarious for two reasons. One, none of their other potential customers really needed the Trusted Computing aspects that Nintendo did, and two, the TSM would be the gateway that would allow the Switch to be jailbroken in the first place.