Intel (INTC -1.60%) has been an industry laggard this year, rising just 5% as the Philadelphia Semiconductor Index rallied 20%. The bulls think Intel can bounce back with its market-leading positions in PC and data center chips, but several Wall Street analysts have become less optimistic about the chipmaker's prospects.

Back in February, Canaccord Genuity analyst Matthew Ramsay downgraded Intel from Buy to Hold, declaring that "shares could remain range bound as margins stagnate" due to peaking PC margins and declining margins at its data center group.

Intel CEO Brian Krzanich.

Intel CEO Brian Krzanich. Source: Intel.

Bernstein analyst Stacy Rasgon downgraded Intel from Market Perform to Underperform in March, noting that it will face "structural headwinds as data center weakens, quality growth becomes more elusive, and competition increases." 

Later that month, Jefferies analyst Mark Lipacis downgraded Intel from Buy to Hold, claiming that its aggressive M&A strategies were making investors "wonder if the market was drifting away from x86" chips. 

Lastly, BofA/Merrill Lynch analyst Vivek Arya recently cut Intel from Buy to Neutral, citing tougher competition from AMD (AMD -5.78%) and NVIDIA (NVDA -3.87%) in its core markets. Investors should generally take these analysts' advice with a grain of salt, but are they right about Intel's dimming prospects? 

The data center slowdown

Intel controls about 99% of the data center CPU market with its Xeon chips. Last quarter, its data center group (DCG) revenue rose 6% annually to $4.2 billion and accounted for 29% of its top line.

That growth sounds decent, but it's well below the 15% compound annual growth rate which Intel claimed that the DCG could achieve back in 2015 -- an estimate which was repeatedly lowered. Intel mainly attributes that slowdown to weaker enterprise spending.

However, the unit's operating profit fell 16% annually to $1.5 billion last quarter, due to higher SG&A expenses and higher product costs from its transition to 14nm chips. Simply put, enterprise customers aren't eager to upgrade their aging servers, so Intel must spend more heavily to develop and promote next-gen Xeons.

Servers in a data center.

Source: Getty.

It's getting harder to make better chips

For several decades, Intel used a two-year "tick-tock" chip production cycle. During a "tick" launch, it introduced a smaller microarchitecture (measured in nm) with a new manufacturing process. During a "tock" launch, those chips are upgraded, but the die size remained the same.

Sticking with that cycle kept Intel ahead of rivals like AMD. But last year, Intel officially changed the tick-tock cycle into a three-stage 30-month cycle dubbed "process, architecture, optimization" -- which extended a chip generation with a final upgrade. That wasn't surprising, since Intel already missed the 24-month tick-tock cycle with its 22nm chips between 2012 and 2014.

In other words, it's getting pricier and tougher for Intel to make smaller (and more power efficient) chips. It introduced its first 14m chip back in Sep. 2014, but its first next-gen 10nm chip won't arrive until late 2017.

Here come the new challengers...

That delay left a wide opening for challengers like AMD, Qualcomm (QCOM -2.53%), and NVIDIA. AMD recently launched Ryzen, a 14nm chip that offers comparable performance as Intel's premium desktop CPUs for about half the price. AMD is pairing these CPUs with its Vega GPUs in its Ryzen APUs for gaming laptops. It also launched its new Epyc CPUs for servers, and Microsoft signed on as one of its first major customers.

Qualcomm also recently launched the ARM-based Centriq 2400, the world's first 10nm server processor. Microsoft also partnered with Qualcomm to launch Centriq-compatible servers. This indicates that Microsoft wants rival chipmakers to break up Intel's near-monopoly of the data center market to lower data center costs.

Lastly, many companies are now pairing NVIDIA's high-powered GPUs with Intel's Xeons for machine learning purposes. Since machine learning operations are more dependent on GPUs than CPUs, data centers could add additional GPUs without upgrading their existing CPUs. NVIDIA's first-mover's advantage in connected cars with its Tegra CPU also bothers Intel -- that's why it agreed to buy Mobileye for $15 billion earlier this year.

So are the analysts right about Intel?

Intel's margins could contract as it tries to spend its way ahead of the tech curve and counter its rivals with cheaper chips. But despite all these headwinds, analysts still expect Intel's revenue and earnings to respectively grow 1% and 5% this year.

Intel's stock is fairly cheap at 15 times earnings, which is well below the industry average of 25 for semiconductor makers. It also pays a decent forward dividend yield of 3.1%. That low valuation and high yield should set a floor under the stock, but it could stagnate until it can show meaningful signs of growth.