Advanced Micro Devices (NASDAQ:AMD) has done extremely well to make a dent in the graphics card business by eating into NVIDIA's (NASDAQ:NVDA) massive lead. But its terrific resurgence has been largely driven by demand from cryptocurrency miners.

AMD's latest Vega GPUs were not powerful enough to match NVIDIA's flagship offerings, but it was able to court cryptominers, and the strategy has worked out perfectly in the short run. However, it won't be long before NVIDIA gets its edge back, thanks to its philosophy of putting gamers first.

Recent revelations, though, indicate that AMD might have jeopardized its chances of giving a tough fight to NVIDIA. The chipmaker was busy pouring resources into another part of the business -- semi-custom chips -- but this move could backfire. Here's why.

An AMD Radeon graphics card.

Image Source: AMD.

AMD makes the wrong move

Forbes says it has gathered from its sources that Sony will be using AMD's Ryzen CPU and Navi GPU architecture in the PlayStation 5, which is expected to hit the market in 2020. AMD has been tied deep into the development of the next-gen Navi GPU architecture, which will be based on the 7-nanometer manufacturing process so that the next Sony console can meet its timeline.

The twist in the tale, however, is that AMD has reportedly developed the Navi architecture keeping the PlayStation 5 in mind, allocating a vast majority of its resources toward this development. As a result, Vega's development was reportedly compromised and failed to match NVIDIA's offerings.

AMD pulled in $2.3 billion in revenue from the enterprise, embedded, and semi-custom segment last year, almost flat from the prior-year period. At the same time, its operating income dropped nearly 46% year over year, to $154 million, because of a seasonal decline in demand for semi-custom chips.

This seasonal weakness isn't surprising as demand for a new console drops after the initial launch happens. For instance, the PlayStation 4 sold 20 million units in 2016, but the number is expected to drop to 16 million this year. Not surprisingly, AMD's operating margin from this segment dropped further, to just 2.6% last quarter, as console demand kept cooling off.

Betting on the wrong horse

Assuming that a successful console moves around 18 million units each year and is on the market for a period of five years, AMD's market would be somewhere around 90 million units. However, last year, PCs consumed a total of 357 million GPUs (including integrated graphics cards), so this is a way bigger market.

As such, spending so much time on developing a GPU for a product that will tail off after an initial spike isn't the best decision that AMD has made.

Metric

2015

2016

2017

2018

PlayStation 4 sales (in millions)

17.5

20

19

16 (estimated)

AMD semi-custom revenue (in $millions)

$2,186

$2,300

$2,305

N/A

Data source: AMD and Sony. N/A = not available 

AMD's semi-custom business started flatlining once the PS4 lost its novelty factor. Of course, the device brought extensive windfall gains for the company at first, because the enterprise, embedded, and semi-custom business supplied just $1.6 billion in revenue in 2013 before the PS4 hit the market.

But this was at the expense of the computing and graphics business, which brought in $3.7 billion in revenue that year. The company's focus on the console segment meant that it lost sight of the GPU market. Not surprisingly, its computing and graphics revenue fell to $2 billion in 2016 due to a weak product lineup before bouncing back to $3 billion last year on the back of cryptocurrency demand.

Now, AMD looks all set to lose its cryptocurrency boost, and the future of the semi-custom business seems cloudy until the next console refresh happens.

Why AMD can't compete with NVIDIA

Now that AMD has reportedly devoted substantial resources toward developing a console-oriented GPU, it will miss out on the high-end PC gaming hardware market. This is because the Navi architecture is expected to be of mid-range spec. This seems like a real possibility, as Navi has originally been developed for a console platform that usually has inferior specs compared to a PC.

By comparison, NVIDIA is looking to push the envelope further by launching its next-generation Turing GPUs later this year. AMD's Vega cards could hardly stand against NVIDIA's two-year-old GPUs when they were launched late last year, so the green army could extend its advantage with a new set of cards.

As such, AMD will be left to compete in the mid-range GPU space, thereby forgoing the fat profits it would have made in the premium segment. In fact, the company's computing and graphics segment was its primary growth driver last year as it delivered an operating profit of $147 million as compared to a $238 million loss the year before.

AMD is betting on a business where its profits are shrinking instead of going for the more prolific GPU business. As such, it isn't surprising to see the company's earnings are expected to grow at a measly compound annual growth rate of just 0.40% for the next five years, as its earnings power will diminish thanks to the lower margin of the semi-custom business and the lack of a cutting-edge GPU product that'll help it compete in the high-end space against NVIDIA.

Harsh Chauhan has no position in any of the stocks mentioned. The Motley Fool owns shares of and recommends Nvidia. The Motley Fool has a disclosure policy.