Track market-moving stories before they get noisy

Moonshot AI's launch of its latest artificial intelligence model, Kimi K3, has shaken global technology markets, reviving memories of the DeepSeek shock earlier this year.

While the model has intensified concerns over the dominance of US AI leaders such as OpenAI and Anthropic, analysts say the broader implications for the AI ecosystem are more nuanced, with several hardware and infrastructure companies potentially emerging as long-term winners.

The Chinese startup claims Kimi K3 rivals some of the world's most advanced AI models despite relying on fewer cutting-edge AI chips, raising fresh questions about the future economics of AI development and spending.

The announcement triggered renewed selling across semiconductor stocks on Friday as investors weighed the possibility that advances in AI efficiency could reduce future demand for expensive computing hardware.

The Philadelphia Semiconductor Index dropped 4% during the session.

"Whatever gap existed between American and Chinese frontier AI just got a lot smaller, and it happened on the exact morning Wall Street was busy convincing itself AI economics don't add up," Mark Malek, chief investment officer at Siebert Financial, wrote following the market reaction on Friday.

Despite concerns surrounding AI chip demand, several investors believe memory manufacturers remain among the strongest positioned companies as AI models continue becoming larger and more capable.

According to Bloomberg, Kimi K3 features 2.8 trillion parameters and supports a one-million-token context window, specifications that require substantially higher memory capacity than previous generations of AI models.

Stanley Tang, senior portfolio manager at Sumitomo Mitsui DS Asset Management, told Bloomberg that memory suppliers should remain among the biggest beneficiaries because the market continues to be dominated by only a handful of companies, including SK Hynix and Samsung Electronics.

Tang added that growing adoption of models such as Kimi K3 is unlikely to reduce overall computing demand.

Instead, broader deployment of agentic AI systems could accelerate hardware consumption over time.

That view is shared by Gary Tan, portfolio manager at Allspring Global Investments.

He told Bloomberg that "the biggest winners will remain the AI infrastructure layer," adding that China's push toward open-source AI would require greater computing resources and continue driving demand for networking equipment and memory chips.

However, whether Nvidia and AMD can maintain the scarcity premium that has driven their valuations is less certain.

While semiconductor shares broadly came under pressure, Mizuho believes concerns surrounding electronic design automation software companies have been overstated.

The brokerage said Kimi K3 strengthens rather than weakens its long-term investment thesis for Synopsys and Cadence Design Systems.

Both companies fell between 8% and 10% last week as investors worried that increasingly capable open-source AI models from China could eventually replace portions of the semiconductor design process.

Mizuho's TMT sector specialist Jordan Klein said those fears were misplaced.

The firm said it "believes this risk is way overblown" and maintained its positive stance on both companies, Investing.com reported.

According to Klein, Kimi K3 functions as a general-purpose AI agent using existing open-source EDA tools such as OpenRoad rather than replacing the underlying software platforms.

He argued that foundation AI models cannot substitute for the deterministic and physically accurate engineering tools required for semiconductor design.

Instead, autonomous AI agents are expected to increase usage of existing EDA software by helping engineers work more efficiently.

Mizuho believes this trend supports its broader "agentic AI engineer" thesis, under which AI helps address the semiconductor industry's engineering talent shortage while expanding monetization opportunities for EDA companies beyond software licensing into engineering productivity, potentially tripling the industry's addressable market over time.

SummaryCompaniesSynopsys informed chipmakers including Samsung, SK Hynix about 'end of life' move, sources saySoftware helps monitor and detect production anomalies during chip productionSynopsys says it is discontinuing select legacy products to focus resources on other higher-value onesSEOUL, July 7 (Reuters) - U.S. chip design giant Synopsys (SNPS.O), opens new tab plans to ​stop offering a suite of manufacturing process control software used by global semiconductor makers, six sources briefed on the matter said, ‌as it seeks to divert resources to higher-margin offerings such as AI design.

Synopsys in April and May informed more than 10 chipmakers including Samsung Electronics, SK Hynix (000660.KS), opens new tab, Kioxia Holdings Corp (285A.T), opens new tab and Qorvo Inc (QRVO.O), opens new tab about the "end of life" move that means Synopsys will not provide future new versions and will only carry out maintenance obligations, two of the sources said.

Make sense of global markets with the Trading Day newsletter. Sign up here.

The ​affected products include the Equipment Engineering System (EES) and Fault Detection and Classification (FDC), a set of automation software that acts as the central nervous ​system of semiconductor fabrication plants to monitor and detect any anomalies before they cascade into costly defects, the two sources ⁠said.

The company has already laid off a few dozen staff, said three of the sources, one of whom added that Synopsys plans to conclude talks ​with each chipmaker on maintenance obligations by July.

Synopsys is discontinuing some legacy manufacturing analytics products to focus resources on the highest-value products, a company spokesperson told ​Reuters in a statement, without naming the products.

The move highlights a changing balance in the semiconductor software industry, where vendors are investing more heavily in AI design technologies while some chipmakers increasingly build manufacturing software in-house.

"While we are discontinuing certain manufacturing analytics products, which are older diagnostic tools not in our customers' critical paths of production, we continue ​to invest in new capabilities in this area of our portfolio and are honoring all existing contractual and support obligations as we take this action,” ​the Synopsys spokesperson said.

The company declined to disclose whether job cuts were involved.

CUSTOMERS LOOK TO DEVELOP IN-HOUSE TOOLSSynopsys began offering the EES product after acquiring semiconductor manufacturing solutions from ‌South Korean ⁠firm BISTel in 2021 for an undisclosed amount.

One of the sources said Synopsys had been wanting to be free of support and maintenance obligations related to IP services and to reallocate engineers to high-margin AI design. Synopsys completed its $35 billion purchase of engineering software firm Ansys, opens new tab in 2025.

That person and a second source said the software's removal risked causing some declines in production yields for chipmakers as the software needed to be constantly maintained, updated and patched.

However, four ​of the other sources said they did ​not expect an impact on ⁠production at major chipmakers.

One of the sources said the decision was also taken partly because enhancing the EES service required chipmakers to share tightly-held manufacturing data. Some clients like Samsung were also developing their own in-house tools, impacting the ​competitiveness of Synopsys' offerings, two sources said.

A Samsung spokesperson confirmed the end-of-life decision and said active discussions were ​underway with Synopsys regarding ⁠the product's sunset. Samsung had established compatible alternatives and there would be "no negative impact on production," the spokesperson said when asked if production yields could decline.

SK Hynix declined to comment. Kioxia and Qorvo did not respond to requests for comment.

Synopsys has for decades been one of the main suppliers of software used in determining ⁠how to ​arrange the tens of billions of transistors that make up chips, which can be 2,000 ​times smaller than the width of a strand of human hair.

In March, Synopsys introduced a technology it said would pave the way toward AI agents taking over many of the tasks in creating ​chips.

Reporting by Cynthia Kim and Hyunjoo Jin in Seoul, Wen-Yee Lee in Taipei and Stephen Nellis in San Francisco; Editing by Brenda Goh and Jamie Freed

Our Standards: The Thomson Reuters Trust Principles., opens new tab