close
close
migores1

5 Super Semiconductor Stocks to Buy Hand Over Fist in 2025

Hundreds of billions of dollars in artificial intelligence spending could hit chip companies next year.

The world’s largest technology companies spend a enormous sums of money building data centers and filling them with graphics processing chips (GPUs) designed for the development of artificial intelligence (AI).

GPUs are built for parallel processing, meaning they can handle large volumes of data and multiple workloads simultaneously, which is essential for training AI models and performing AI inference. Based on recent financial reports:

  • Microsoft spent $55.7 billion on capital expenditures (capex) in fiscal 2024 (ended June 30), most of which went to AI data center infrastructure and chips.
  • Amazon spent $30.5 billion on investments in the first half of 2024, mostly related to AI.
  • Alphabet spent $25 billion on AI investments in the first half of 2024.
  • Meta platforms plans to spend between $37 billion and $40 billion on AI investments for the full year of 2024.
  • Oracle spent $6.9 billion on AI capex in fiscal 2024 (ended May 31).

Microsoft, Amazon, Meta and Oracle have all explicitly said they plan to spend even More money in the following year. That means opportunity for the semiconductor industry, so here are five top stocks to buy in 2025.

A digital rendering of computer chips, with one labeled AI.

Image source: Getty Images.

1. Nvidia

When it comes to data center GPUs, Nvidia (NVDA 3.37%) is the undisputed leader. Its H100 GPU set the benchmark for the AI ​​industry last year, and the company is now preparing to deliver an entirely new generation of chips based on its Blackwell architecture.

Blackwell-based systems such as the GB200 NVL72 will perform AI inference up to 30 times faster than equivalent H100 systems. Additionally, individual GB200 GPUs are expected to sell for between $30,000 and $40,000 each, which is in line with what data center operators initially paid for the H100. In other words, Blackwell paves the way for a substantial improvement in cost efficiency for developers, who normally pay for compute capacity by the minute.

Nvidia CEO Jensen Huang expects Blackwell GPUs to start contributing billions of dollars to the company’s revenue in the last quarter of fiscal 2025 (which runs from November to January), and shipments will increase from there .

Nvidia is on track to generate an estimated total revenue of $125.5 billion in fiscal year 2025, representing a 125% year-over-year increase. Its stock isn’t cheap, but it trades at a reasonable forward price-to-earnings (P/E) ratio of 29.1 when measured against the company’s forecast earnings per share for fiscal 2026.

Simply put, investors who are willing to hold Nvidia stock for at least the next 18 months are likely paying a reasonable price today.

2. Micron technology

Micron technology (MU 1.97%) is a leading supplier of memory and storage chips for data centers, personal computers and smartphones. In AI data center settings, memory chips complement GPUs by storing data in a ready state so it can be instantly called upon for training and inference. As AI requires a lot of data to learn and improve, the demand is increasing for more memory capacity.

Micron’s latest 36 gigabyte (GB) HBM3E (high-bandwidth memory) drives for the data center offer up to 50% more capacity than any competing product today, while consuming 20% ​​less power little energy. It has been selected to power Nvidia’s H200 GPU and potentially its Blackwell GB200 GPUs. In fact, Micron is now completely out of HBM3E by 2026.

Beyond the data center, every Tier 1 manufacturer of Android-based smartphones uses Micron LP5X DRAM memory. Many of them have launched AI-enabled devices this year with minimal memory capacity requirements doubling compared to their non-AI predecessors last year. There’s a similar trend in personal computing, with most AI-enabled PCs launching with a minimum DRAM capacity of 16GB, compared to 12GB last year.

Higher memory requirements translate directly into higher revenue for Micron. In its recent fiscal 2024 fourth quarter (ended Aug. 29), the company’s revenue rose 93% year-over-year to $7.7 billion, and it expects more to come.

3. Axcelis Technologies

Axcelis Technologies (ACLS -0.97%) is not a semiconductor manufacturer. Instead, it makes ion implantation equipment, which is critical to the manufacturing process for central processing units (CPUs), memory chips, and power devices that regulate the flow of electricity in high-current applications.

AI data centers have become a new opportunity for power device manufacturers (and Axcelis equipment) because they consume substantial amounts of energy. Some of Axcelis’ customers have begun using MOFSET (metal oxide semiconductor field-effect transistor) power devices with a silicon carbide chemistry that is more robust and thermally efficient than traditional chemistries of silicon. Silicon carbide power devices require more implants, which is a direct tailwind for Axcelis’ business.

In addition, Axcelis is benefiting from the growing demand for high-capacity memory chips in the data center, computers and smartphones. As memory chip makers expand production to meet this demand, they will need more Axcelis equipment, and the company has already started building inventory in anticipation of a strong 2025.

In fact, Axcelis’ guidance suggests that 2025 could be the strongest year in the company’s history, with a potential record $1.3 billion in revenue on the cards.

4. Broadcom

Broadcom (AVGO 0.72%) is a multi-faceted AI organization with presence in semiconductors, hardware, cybersecurity, cloud software and more. The semiconductor business is in the spotlight for investors right now, as many of its products face significant demand from companies building AI infrastructure.

Broadcom makes AI accelerators (chips for data centers) for hyperscale customers (which typically include Microsoft, Amazon, and Alphabet). During the recent third quarter of 2024 (ended August 4), the company said business grew three and a half times over the year-ago period. Similarly, it saw a fourfold increase in sales of its Tomahawk 5 and Jericho3-AI Ethernet switches for the data center. They regulate how fast data travels from one point to another.

In the third quarter, Broadcom is expected to generate total revenue of $51 billion for fiscal 2024, with $11 billion attributable to AI. However, following the strong results we just highlighted, the company revised those numbers to $51.5 billion and $12 billion, respectively.

Broadcom is currently knocking on the door of the exclusive $1 trillion club, home to only six US tech giants.

5. Advanced Micro Devices (AMD)

Advanced microdevices (AMD 1.92%) is one of the leading chip suppliers in the consumer electronics industry. Its hardware can be found in anything from SonyHis PlayStation 5 to infotainment systems from adzehis electric vehicles. However, the company is now trying to compete with Nvidia in the data center segment with its new MI300 GPUs.

The MI300 was designed as an alternative to Nvidia’s H100 and appeals to top customers such as Oracle, Microsoft and Meta platforms. However, with Nvidia set to release the Blackwell lineup, AMD is already preparing to launch a faster GPU called the MI350 next year. It will be based on a new architecture called Compute DNA (CDNA) 4, which could offer 35 times more performance than CDNA 3 chips like the MI300. AMD intends this architecture to compete directly with Blackwell.

AMD is already the leader in AI chips for the personal computing segment, with an estimated 90% market share. It recently launched the Ryzen AI 300 series for notebooks, which features the fastest Neural Processing Unit (NPU) in the industry. The company says more than 100 platforms will launch with those chips from top PC makers like Asus, Acer, HP Inc.and many others.

AMD has a staggering 115% year-over-year growth in its data center revenue and a 49% increase in revenue from the client segment (which houses its Ryzen AI chips) in the second quarter of 2024. But the AI ​​revolution it’s just warming up, so the best may be yet to come.

Related Articles

Back to top button