If you take those kind of statements at face value, XPoint is behind.
Whatever claims Intel made, WD made the claims that their 3D ReRAM has better perf, better costs and much better scaling with production starting in second half 2018.
As for Intel, they said NAND will be profitable in the second half of 2017 and the segment overall in late 2018.This means XPoint would still be bellow break even at that time- so commercially viable only 2019 at best. This is likely also a factor of DRAM pricing. If DRAM prices go back to first half of 2016 levels by then, it puts pressure on XPoint ASPs. NAND pricing can put some pressure on it too.
Intel and Micron are first to market ,that's clear but from a technology point of view, they could very well be far behind. At the very least the point resistion made about costs and scaling does stand, repeating every step for each layer is not exactly the most efficient way to manufacture and scale 3D memory.
3D NAND is TLC, so its reliability is pretty terrible (retention, endurance). 3D XPoint's main advantage is speed (faster than flash but slower than DRAM) but its 3D stacking scheme is much more expensive than that of 3D NAND. 3D XPoint reliability doesn't appear much better than that of 3D NAND either.
Intel is spending a lot of money on NVM. Intel is also the only company losing money in NVM. and they are losing a lot in a one of the best markets in history.
Intel has products out on 3d XPoint. It is clear that the endurance is 5x that of NAND best case (MLC NAND is >3000cycles, SLC >10K cycles). it is also faster than NAND but 10x slower than DRAM. So what is replaces and whether or not Intel has a large lead is debateable. No one expects XPoint to replace 3D NAND
Also Intel continues to sell in SSDs and sell to WDC and Intel sales of component NAND have always been very small.
Summary: Intel is growing the SSD business nicely and introducing fast NAND (Crosspoint) while losing a lot of money.
"First, Intel's memory group [NSG] has announced $866 million in revenue in Q1 2017, which is 55% more than $577 million in Q1 2016."
Pay attention and do the math. Look at how bit output grows for the industry , account for the fact that Micron and Intel are above that average this cycle ( they were bellow average in the previous cycle) and then account for NAND pricing.
Intel's CAPEX for NAND is due to them taking some production outside the JV not because there is a so called ramp. This bit is interesting, if the press would ask why but the press falls for the official nonsense because the press is a marketing vehicle nowadays. Look at clean room area for 2D vs 3D NAND, terms of the JV, Intel's (idle) fab capacity and maybe you find an answer.
MIcron NV revenue (approx in millions) in the last 4 reported quarters 1400, 1270, 1000 , 900.
Vs Intel's 866, 816, 649, 554.
Revenue is a factor of volume, price and mix ofc but even when ignoring mix, see how both scale by a bit over 55%? It's the process transition (in this case 2D to 3D) and NAND pricing, no evidence of any ramp on top of it, at all.
The main purpose of this article was to get more clarity on Intel's memory business. And it presents facts, not just opinion.
First, Intel's memory group [NSG] has announced $866 million in revenue in Q1 2017, which is 55% more than $577 million in Q1 2016.
Second, Intel is raising its capex both on flash and 3D Xpoint memory technologies. It's also a fact, widely reported in trade media.
Third, Intel and Micron are years ahead with 3D Xpoint as a flash replacement, compared to Samsung, Hynix and Toshiba. The article notes, however, that things aren't very clear on where 3D Xpoint stands in terms of readiness. We will most likely know later this year.
Look at Intel's NAND revenue, assume that global NAND bit output grows by some 40% per year, assume that Micron and Intel are a bit above that this cycle and then factor in how NAND prices have evolved over the last few quarters. Now show us any evidence at all of anything unusual and a so called ramp. It's a white lie for investors since Intel now has a reporting segement for NVM. Going forward, Xpoint will mask the pseudo-ramp for NAND so folks will keep thinking it's there - standard operating procedure for corporations like Intel.