VC Funding Shifts to Pre-IPO Mega-Rounds as Early-Stage Deals Slide Global venture capital investment is increasingly flowing to proven companies in a winner-take-most pattern, and the concentration is expected to sharpen this year around mega-rounds for firms nearing an initial public offering. According to VC analytics firm THE VC, investment in unlisted startups and small and midsize companies at the seed-to-Series A stage totaled 1.919 trillion won last year, down 41.17% from a year earlier. The number of deals fell even more sharply, down 67.65% to 847. Later-stage funding rose. Series B-to-C investment reached 3.2732 trillion won, up 6.06%, while Series D-to-pre-IPO funding climbed 20.74% to 1.3802 trillion won. In late-stage rounds, the number of deals also increased 34.92%. The share of funding by stage shifted as well. Early-stage rounds accounted for 29.2% of total investment, down 9.8 percentage points from the 39% range in 2024. Midstage and late-stage rounds rose to 49.8% and 21%, up 5.3 and 4.5 points, respectively. THE VC said pre-IPO investment increased noticeably, attributing it to signs of recovery in the IPO market, including broader improvements in the number of new listings and IPO demand indicators. With expectations rising for South Korea’s stock market alongside talk of reaching a KOSPI 5,000, THE VC projected that the shift of money toward pre-IPO companies will accelerate. In South Korea, AI startups such as FuriosaAI, Upstage and Superb AI are being discussed in the investment banking industry as potential IPO candidates in 2026. The tilt toward big late-stage deals is also a global trend. Crunchbase estimated that global VC investment in AI totaled US$211 billion last year, up 85% from a year earlier and above the 2021 peak. Much of the money went not to seed-to-Series A startups but to established companies such as OpenAI. Last year, OpenAI, Anthropic, Scale AI, xAI and Project Prometheus each raised mega-rounds of at least US$5 billion. Together, the five companies raised nearly US$84 billion, about 20% of global VC investment. U.S. media outlets including The New York Times have reported that OpenAI and Anthropic are in IPO preparation. OpenAI’s valuation, estimated at US$500 billion, could approach US$1 trillion after a listing, according to forecasts cited in the reports. Anthropic’s valuation, estimated at US$183 billion, could rise to as much as US$350 billion after an IPO. Capital concentration was also pronounced. About 60% of global investment capital last year went to 629 companies that raised at least US$100 million, and more than 30% flowed to 68 companies that raised at least US$500 million. A South Korean VC official said, “Starting a business will get harder, and more money will flow to companies that have begun making money.” The official added, “VCs have more cash than ever, but the mood has shifted from taking on challenges to investing in profitability.”* This article has been translated by AI. 2026-01-28 18:33:26

Nvidia AI chip for self-driving to test on Mercedes Benz in US roads Q1 LAS VEGAS, January 06 (AJP) -Nvidia unveiled a next-generation “reasoning AI” for autonomous driving at CES 2026 on Monday, signaling a deeper push into the self-driving vehicle market with real-world deployments planned as early as the first quarter of this year. At a press conference at the Fontainebleau Live Theater in Las Vegas, Nvidia CEO Jensen Huang introduced the technology, dubbed Alpamayo, describing it as “the world’s first self-driving AI that thinks and reasons.” “The ChatGPT era of physical AI has arrived,” Huang said onstage. “Machines have begun to understand the real world directly and to reason and act on their own.” Huang said Alpamayo is designed to address autonomous driving’s so-called “long-tail” problem — rare, complex scenarios that are difficult to anticipate through conventional data collection alone. Unlike traditional systems that react primarily to sensor inputs, Alpamayo reasons step by step, Huang said, and can explain why it chose a particular action. “Alpamayo not only takes sensor input to control steering, braking and acceleration, it also reasons about what action to take and tells you why,” he said. “It can respond safely even in complex long-tail scenarios.” The model is a vision-language-action system built on Nvidia’s physical AI platform, Cosmos, which the company describes as a world foundation model. Nvidia said Alpamayo was trained using human driving demonstrations and trillions of miles of AI-generated data. During the presentation, Huang showcased a live demonstration of a vehicle equipped with Alpamayo navigating downtown San Francisco, verbally explaining its decision-making process in real time. “The long tail of driving can’t be solved by collecting every scenario in the real world,” Huang said. “But it can be solved by breaking it down into smaller, normal situations and reasoning through them.” A video shown onstage featured an AI-powered Mercedes-Benz vehicle driving through San Francisco traffic while a passenger sat behind the steering wheel with their hands resting in their lap. “It drives so naturally because it learned directly from human demonstrators,” Huang said. “In every single scenario, it reasons about what it’s going to do and why.” Nvidia said it plans to release Alpamayo as open source, with the underlying code made available on the machine-learning platform Hugging Face, allowing autonomous-vehicle researchers to access and retrain the model. Several companies and research groups have already lined up to use Alpamayo for Level 4 autonomous driving development, including Lucid Motors, Jaguar Land Rover, Uber and Berkeley DeepDrive. “Robotaxis will be the first to benefit,” Huang said, adding that Mercedes-Benz vehicles equipped with Alpamayo are expected to begin operating on U.S. roads in the first quarter. Nvidia also said it is planning a robotaxi pilot service in 2027 with partners including Uber. The company declined to name additional partners or specify locations for the rollout. “Our vision is that someday, every single car, every single truck, will be autonomous,” Huang said. 2026-01-06 13:29:17

[[CES 2026]] Nvidia's Jensen Huang, AMD's Lisa Su to outline AI strategies LAS VEGAS, January 05 (AJP) - Nvidia CEO Jensen Huang, AMD CEO Lisa Su and Lenovo Chairman Yang Yuanqing are set to take the stage at this year's CES, the world’s largest technology trade show, as artificial intelligence remains the event’s central theme. The three executives are expected to outline industry trends and showcase their companies’ strengths across AI infrastructure, physical AI and AI inference during keynote appearances at the show in Las Vegas. After last year’s heavy focus on AI agents and infrastructure, attention at CES 2026 is expected to shift toward inference, physical AI and industrial applications, industry officials said. Nvidia said on Monday that Huang will deliver a special address on Jan. 5 local time, a day before the official opening of CES, at the Fontainebleau hotel in Las Vegas. Huang is expected to present his outlook on the AI industry and Nvidia’s strategy, with market attention focused on the performance and competitiveness of the company’s next-generation graphics processing unit, known as Rubin. He is also scheduled to meet major global media outlets on Jan. 6, where he may discuss plans related to high-bandwidth memory supplies with South Korea’s Samsung Electronics and SK hynix. Nvidia has been expanding its supply chain amid surging demand for AI chips used in data centres, and details of potential partnerships could be disclosed. AMD said Su will deliver a CEO keynote on Jan. 5 local time, where she is expected to outline future AI solutions and the company’s strategy to compete with Nvidia. With AMD emerging as a leading challenger in the AI GPU market, attention is focused on whether it will unveil next-generation AI chips. AMD has also been targeting the AI personal computer market through updates to its Ryzen processor lineup, and new Ryzen products are expected to be announced at CES. Lenovo’s Yang is expected to focus on integrating AI features into IT devices such as personal computers and mobile products. The company plans to introduce new AI-enabled laptops and tablets aimed at improving productivity and user experience. Siemens Chairman Roland Busch is also scheduled to speak at the show. Busch will deliver a keynote on Jan. 6, highlighting industrial AI and automation solutions and how AI can drive innovation across manufacturing, energy and transportation. 2026-01-05 09:57:51

KAIST nuclear engineering professor named head of Korean Nuclear Society SEOUL, December 31 (AJP) - Choi Sung-min, a professor of the Department of Nuclear and Quantum Engineering at the Korea Advanced Institute of Science and Technology (KAIST), will assume office as new president the prestigious Korean Nuclear Society, effective Jan. 1. Choi earned his bachelor’s and master’s degrees in nuclear engineering from Seoul National University, before completing a master’s degree and a doctorate at the Massachusetts Institute of Technology (MIT). He later served as a visiting researcher at the U.S. National Institute of Standards and Technology (NIST) and is currently a professor at KAIST. He has held a number of senior roles in the nuclear field, including chair of the board of the Korea Institute of Nuclear Safety, president of the Asia-Oceania Neutron Scattering Association, president of the Korean Neutron Beam Users Association, head of the Center for Advanced Nuclear Technology Research, and director of the Nuclear Basic Joint Research Institute. In a statement marking his appointment, Choi described nuclear power as “a key national asset” capable of meeting rising electricity demand in the era of artificial intelligence while supporting carbon neutrality goals and energy security. He added that the society would present future energy solutions grounded in scientific evidence and objective data. 2025-12-31 09:26:34

South Korea's stock market rides early wave of AI-driven supercycle SEOUL, November 05 (AJP) - South Korea’s stock market appears to be entering the early stages of what analysts are calling an artificial intelligence–driven “supercycle,” as investors pour into semiconductor, robotics, and infrastructure shares amid growing optimism about the country’s role in the global AI boom. The rally gained momentum following recent international events such as the APEC summit and Nvidia’s AI conference, both of which underscored the long-term potential of the technology. The benchmark KOSPI index briefly surpassed 4,200 on Nov. 3, hitting an all-time high, before easing to close at 4,121.74 on Nov. 4, down 100.13 points, or 2.37 percent. Even so, many analysts see the pullback as a pause in a broader upward trend. Much of the enthusiasm centers on South Korea’s AI ecosystem, particularly in semiconductor and data center infrastructure. Shares of SK hynix have surged nearly 183 percent this year, while Samsung Electronics is up 83 percent and Naver has gained about 41 percent — a reflection of booming demand for AI computing capacity. The rise of AI data centers has also fueled gains among power and equipment suppliers. HD Hyundai Electric’s stock has soared about 143 percent on demand for high-voltage transformers, while Iljin Electric has advanced 118 percent. Government-backed domestic data center projects have reached record order backlogs, feeding investor optimism that the momentum will continue. In the “physical AI” sector, companies linked to robotics and autonomous systems are also reaping benefits. Doosan Robotics’ shipments of collaborative robots are expected to climb 50 percent this year, lifting its shares by more than 84 percent. Rainbow Robotics, known for its AI-powered robotic arms, is up about 80 percent, while Hyundai Autoever has gained nearly 58 percent on hopes for its autonomous driving software. Analysts remain bullish on the outlook. Kim Byung-yeon, an analyst at NH Investment & Securities, projected that the KOSPI could reach 5,000 by 2026, citing structural shifts in industrial competitiveness rather than low interest rates. KB Securities offered a similar forecast, also setting a 2026 target of 5,000 and estimating that global AI data center investments could exceed $1 trillion by 2028. That optimism stands in contrast to mounting concern in the United States about an “AI bubble,” where companies such as Google and OpenAI face scrutiny over soaring R&D costs and potential overinvestment. According to a recent Bloomberg report, U.S. technology giants could collectively spend more than $1 trillion on AI development by 2025 — a pace that some analysts say risks overheating. South Korea’s approach, by contrast, has emphasized state-backed infrastructure and “green AI” initiatives designed to mitigate energy strain and foster long-term stability. Unlike in the U.S., where investor enthusiasm often outpaces corporate performance, South Korea’s AI boom has been accompanied by measurable earnings growth. SK hynix reported a 61.9 percent year-on-year increase in third-quarter operating profit, to 11.38 trillion won ($8.2 billion), driven by sales of high-bandwidth memory chips. Samsung’s shipments of AI server processors rose 1.8 times from the previous quarter, while Naver’s AI cloud business expanded its utilization rate beyond 20 percent, strengthening its recurring revenue base. Analysts say the country’s AI momentum is rooted less in hype than in hardware for now — and in the structural shifts that could redefine South Korea’s role in the global technology supply chain. * This article, published by Aju Business Daily, was translated by AI and edited by AJP. 2025-11-05 09:46:22

INTERVIEW: South Korean start-up bets on AI to transform learning Long before the recent surge in artificial intelligence triggered by OpenAI in 2023, Ellis Group had been quietly embedding AI tools into education. Founded a decade ago by a team of researchers from the Korea Advanced Institute of Science and Technology, the company now counts more than 9,500 institutions and 2.75 million users on its AI education platform, EllisLXP. The Seoul-based firm has become a leader in South Korea’s AI education sector, providing software for universities, corporations and government agencies. Its cloud services, EllisCloud, are also used by more than 4,000 organizations for AI research. CEO Kim Jae-won says the company’s next chapter will focus on global expansion. In an interview, Kim discussed Ellis Group’s origins, its ambitions and what sets it apart in the fast-growing market for AI-powered learning. Q. How would you describe Ellis Group? A. Ellis Group provides B2B software-as-a-service solutions, including EllisLXP for learning and EllisCloud for research and infrastructure. Founded in 2015, our platforms are widely used across South Korea by corporations, universities and public agencies. EllisLXP alone has more than 9,500 institutional partners and 2.75 million users. EllisCloud supports AI research at over 4,000 organizations. This year, we also introduced a broader infrastructure-as-a-service product, which includes modular data centers, servers and cloud platforms to help companies transition into AI. Q. What are your flagship products? A. EllisCloud is anchored by the Ellis AI PMDC, a modular data center designed for startups and research facilities. It delivers high-performance GPU and NPU resources and allows users to scale clusters securely. It also earned CSAP IaaS certification earlier this year. EllisLXP, our education platform, is an all-in-one solution with tools such as AI-generated quizzes, voice-recognition subtitles and a virtual tutor, “AIHelpy,” which adapts to learners’ age and cultural context. Q. What led you to start the company? A. While conducting research at KAIST, I saw how difficult it was to deliver large-scale computer education. My colleagues and I developed a prototype for AI-driven grading to address that challenge. In 2015, Kim Su-in, Park Jung-kook and I founded Ellis Group with the vision of using AI to improve education and nurture talent for South Korea’s AI future. Q. What differentiates Ellis Group in the AI education space? A. We focus on lifelong learning. Our solutions are used in schools, corporations and government agencies, offering project-based learning that ties directly to real-world challenges. Unlike many competitors, we combine infrastructure, AI models and educational content under one umbrella. With both SaaS and IaaS certified under CSAP, we are uniquely positioned to support large-scale public and private initiatives. Q. How do you see the company’s growth trajectory? A. We anticipate rapid growth in our cloud services, supported by expanding adoption of both EllisLXP and EllisCloud overseas. Q. What do you look for in new employees? A. We value people who are eager to grow and who care about diversity and social impact. Our mission requires solving complex, global problems with AI, and we want individuals ready to contribute to that effort. Q. What is the roadmap for Ellis Group? A. Education and infrastructure will remain at the center. We plan to extend AI education to more schools and companies while strengthening our cloud offerings. We are continuing development of Ellis AI PMDC and preparing for global expansion. Q. Any further comment? A. Ellis Group hopes to play a meaningful role in South Korea’s development in the AI era, through both education and infrastructure. We are grateful for the support we’ve received and excited for what comes next. * This article, published by Aju Business Daily, was translated by AI and edited by AJP. 2025-10-02 09:02:01