Intel has finally admitted that many of us saw a long time ago, that they have "lost their way", which is what Intel's chairman recently said. He's of course referring to the boom in the smartphone and tablet market, which has caught Intel by surprise, and completely not ready to tackle that market with their old Atom chip that was designed for netbooks, and used to draw 15-20W of power in the early days.
Intel's Atom was indeed more powerful than the early 400 Mhz ARM11 chips we saw in the first iPhones and other touchscreen smartphones, which allowed Intel to only focus on dropping power consumption, instead of raising performance (which is something they never wanted to do in those days anyway, because they didn't want Atom to compete with their Core line of chips).
Unfortunately for Intel, ARM chips kept improving at a much faster rate in performance, than Intel was improving Atom in power consumption, so by the time Intel Atom reached parity in power consumption, ARM's chips have already surpassed it in performance, now being about a whole generation ahead, in both CPU and GPU performance, as as Intel is ahead with its process node (22nm vs 28nm).
You can see here how Intel's latest Bay Trail can barely match the 2012 CPU and GPU's (Adreno 320 is actually a 2012 GPU, first released in S4 Pro – Adreno 330 is twice as powerful):
Even at this point in the mobile industry, Intel still doesn't have a proper competitor for smartphones, which is really the biggest part of the mobile market, as it's close to reaching 1 billion new units per year. Intel's smartphone chip, Merrifield, which is based on the new Silvermont architecture, isn't scheduled for being shipped until the first half of 2014, and going by the little success Intel has had in the market so far, it's probably not going to arrive in any major smartphone model either.
Intel is preparing 2 new chips for the second half of next year, Broxton for high-end (probably as a replacement for Bay Trail or Merrified – it's not clear whether it's a smartphone chip or a tablet one), and one called Sofia, which is targeted at the low-end and most likely a replacement for Intel's Lexington, which was largely a big failure on Intel's part because of its very poor performance.
Broxton will have to compete with the 20nm Cortex A57, Samsung's own ARMv8-based custom CPU, and Qualcomm's 64-bit successor to the recently announced Snapdragon 805. Sofia, on the other hand, will have to compete with quad-core and 8-core 64-bit Cortex A53 chips at the mid-range and low-end. It's uncertain how Intel will fare with these CPU wise, but GPU-wise will likely remain behind Tegra 5 and Mali T760, unless Intel has found a way to quadruple the performance of Bay Trail's GPU within only one year (which sounds very unlikely.
Until then Intel will surely hype up its products as much as possible, but if I've learned anything over the years, there's quite a large gap between what Intel announces a whole year ahead about its products, and how they actually fare against the competition when they arrive.
The most recent such case being when Intel promoted Iris Pro graphics heavily, making everyone think that if they buy Haswell laptops, they will have 2-3x the GPU performance of the IVB laptops, when in fact Iris Pro was only ever destined for very high-end machines (it's only available in the Macbook Pro with Retina right now), and not for mainstream Haswell laptops, which only had ~20-30 percent graphics improvement over last year's IVB laptops, on average. That's why I'm usually very skeptical about Intel's claims these days, and I recommend waiting until their chips actually launch and get reviewed, before making up your mind about them.