Intel’s stock prices have plummeted 8% today, following the company’s revelation of significant operating losses in its fledgling foundry business.
In 2023, Intel’s foundry business reported an operating loss of $7 billion, with sales amounting to $18.9 billion. This marks a larger loss compared to the $5.2 billion loss reported in 2022, despite higher sales of $27.5 billion during that period.
For the first time, Intel has revealed the revenue figures specifically for its foundry business. Traditionally, Intel has engaged in both chip design and manufacturing, reporting overall chip sales to investors.
In contrast, other American firms like NVIDIA and AMD focus entirely on chip design, outsourcing manufacturing to Asian foundries, such as TSMC or Samsung.
Contrastingly, Intel once held the mantle of technological leadership in the semiconductor sector. However, a prolonged stagnation in the development of their 10nm fabrication process over half a decade essentially sealed their fate as a market leader. Meanwhile, competitors like TSMC continued to advance, maintaining their targets.
In an attempt to reassure investors about the future of their semiconductor sector, Intel embarked on a complete rebranding of their manufacturing roadmap in 2021. They equated their 10nm Enhanced SuperFin processing node with TSMC’s 7nm process, labeling it “Intel 7” moving forward, while their 7nm process was relabeled as “Intel 4.”
This rebranding effort, though convoluted, aimed to appeal to investors because smaller numbers as better. However, Intel’s struggles have only become more apparent despite the rebrand.
The introduction of Alder Lake marked the initial transition to their 10nm/Intel 7 process, after five years remaining stagnate on their 14nm processing node, but fierce competition from rivals like AMD has underscored Intel’s CPUs as being notably inefficient.
Subsequent rebrands such as Raptor Lake and the 14th Generation Refresh have failed to rejuvenate Intel’s leadership and efficiency concerns.
The latest offering, the Intel Core i9-14900KS, notably consumes a staggering 508 watts in multi-threaded workloads, doubling the power consumption of AMD’s Ryzen 9 7950X and is 6.59 times that of the AMD Ryzen 7 7800X3D, which remains the fastest processor for gaming workloads.
For fanboys loyal to the blue brand, Intel’s strides with the 12th and 13th Generation processors indeed brought notable boosts in performance, particularly in terms of Instructions Per Clock (IPC). However, this progress came at the expense of regression in performance for certain applications, owing to its hybrid design.
This design incorporates core complexes consisting of both high-performance cores and efficiency cores. Unlike AMD’s unified approach with compact cores, Intel’s design does not feature cores that share the same Instruction Set Architecture (ISA).
Intel’s Hybrid design incorporates cores with varying levels of IPC (Instructions Per Clock). Compounded by the fact that E-cores operate at a significantly lower frequency, they are not ideally suited for managing sensitive workloads.
To address this, Intel introduced a software solution called the Application Performance Optimizer.
This software artificially boosts the frequency of these efficiency cores when needed, enhancing gaming performance by up to 18%. However, this feature is exclusively available on the 14th Generation of Core processors, despite them being identical to the previous 13th Generation and sharing the same “Gracemont” efficiency cores as the 12th Generation.
Looking ahead, Intel’s latest processing node, Intel 4, has been met with disappointment. The anticipated true 14th Generation lineup, now named Intel Core Ultra, is exclusively designed for mobile devices under the name “Meteor Lake.”
These processors not only exhibit a regression in clock speed compared to previous generation hardware but also seem to offer inferior IPC (Instructions Per Clock) performance compared to the previous generation. Various sources, including NotebookCheck and notably Phoronix, have reached similar conclusions, highlighting that Meteor Lake is significantly outperformed in terms of both performance and efficiency by AMD Zen 4 processors.
This outcome is far from satisfactory for a company striving to rebuild its faltering image and reputation.
While initially planned for a widespread desktop release, Meteor Lake failed to materialize as Intel scrapped those plans in favor of a large-scale rebranding effort for the 13th Gen Raptor Lake CPUs.
Now, Intel aims to counter AMD with “Arrow Lake,” essentially a refined version of Meteor Lake. Intended to leverage Intel’s 20A process (2nm), Meteor Lake on 7nm did improve upon Intel’s lackluster efficiency compared to previous generation 10nm processors, but it still falls far behind AMD’s offerings.
Adding to the frustration is the fact that Arrow Lake is essentially a tweaked iteration of Meteor Lake, serving as little more than a beta test, similar to the role played by the 12th Gen Alder Lake CPUs.
Leaked Intel slides and reports from sources like Igor’s Lab suggest that Arrow Lake processors will only deliver a modest 6-21% improvement in multicore performance over Raptor Lake Refresh, with median single-core performance projections showing a mere 9-11% increase.
This is disappointing, considering Arrow Lake is expected to utilize what’s believed to be a vastly superior processing node, falling well below performance expectations and will likely be demolished by AMD’s Zen 5 architecture if such metrics are to be believed.
It seems Intel is pivoting towards becoming another TSMC customer, evident from their utilization of TSMC silicon in their ARC graphics cards, which have been criticized for their lackluster performance. The self-proclaimed “world’s largest chip manufacturer” is now turning to TSMC silicon for its upcoming tile-based CPUs as well.
Current-generation Meteor Lake CPUs are being built using a combination of Intel 4 and TSMC’s 6nm and 5nm nodes for its CPU tile, SoC tile and integrated graphics portions respectively.
In recent years, Intel’s actions as a company have largely been perceived as lackluster or even embarrassing failures. Under CEO Patrick Gelsinger’s leadership, Intel has been pitching investors on a strategy where it continues to manufacture its own processors while also venturing into an external foundry business to produce chips for other companies.
Curiously, Intel has expressed a desire for competitors like AMD to become customers, potentially building their products on Intel’s own silicon, which is fucking hysterical, because why would AMD opt to use inferior silicon and give their main competitors access to their crucial IP?
Intel’s position as one of the few U.S. companies engaged in cutting-edge semiconductor manufacturing within the country played a significant role in its recent securing of nearly $20 billion in funding under the CHIPS and Science Act.
This funding is supposedly essential for Intel to maintain competitiveness in the relentless semiconductor industry race. However, Intel is utilizing government grants, likely funded by taxpayers, to secure wafers from its main rival, TSMC.
It seems Patrick’s assertion about Intel’s recently spun-off Foundry sector becoming the world’s second-largest foundry by 2030 may indeed rely heavily on significant government funding. Without such support, there’s a high likelihood of being overtaken in node capabilities even by Samsung.
Intel’s actions, including their partnership with TSMC for silicon, suggest an acknowledgment of their fabrication process is second rate at best.
Currently, much of Intel’s foundry revenue stems from its own operations, as stated by the chipmaker. Moreover, Intel has restructured its Products division to account for foundry expenses as if it were a “fabless” company.
In this newly organized structure, which primarily focuses on processors for PCs and servers, Intel reported $11.3 billion in operating income on $47.7 billion in sales in 2023.
On Tuesday, Intel announced its expectation for its foundry’s losses to reach a peak in 2024 and then gradually break even by the midpoint between the current quarter and the end of 2030.
The company had previously disclosed that Microsoft would utilize its foundry services, with $15 billion in foundry revenue already secured, however the prospect that even Intel doesn’t foresee its foundries producing a profit in half a decade is more than concerning.
During an investor call on Tuesday, Gelsinger emphasized the significant earnings growth that Intel Foundry is poised to drive over time, stating that 2024 represents the low point for foundry operating losses.
In a promotional video, Intel attributed much of the lack of profitability in its foundry business to the “weight of past decisions.” Additionally, Gelsinger mentioned the company’s historical sluggish adoption of Extreme Ultraviolet Lithography (EUV), a key technology for manufacturing the most advanced chips, as a contributing factor to the challenges faced by the foundry business.
Intel intends to invest $100 billion in constructing or expanding chip factories across four U.S. states. The success of its business turnaround strategy hinges on convincing external companies to utilize its manufacturing services, however, from my perspective, I find it difficult to envision the company reclaiming its position as an industry leader. Their best chance may lie in unforeseen geopolitical shifts such as a Chinese invasion of Taiwan.
Alternatively, they could focus on aggressively recruiting talent from TSMC to bolster their chances.
