SoftBank Group has been among the top performers on Tokyo’s Prime market this year, surging on the back of a US tech rally powered by an artificial intelligence boom.

The conglomerate’s increasing AI investments have fueled hopes for lucrative returns, but some investors are still mulling whether it’s worth the bet.

Shares in SoftBank are up more than half this year, gaining steam over the last few months and outperforming the benchmark Nikkei Stock Average’s 8.5% gain. SoftBank touched an all-time closing high of JPY 16,705 (USD 113) on Aug. 18, helping push the Nikkei average to a record high last month, before tumbling to the 14,000 level.

Tomoichiro Kubota, a senior market analyst at Matsui Securities in Tokyo, said the stock’s more recent weakness reflected concerns among retail investors that its rally was due for a pause.

“SoftBank’s share price doubled in a short period of time as expectations and hopes for its future growth climbed very quickly,” he said. “Long-term institutional investors operate differently, but for Japanese retail investors, it was an opportunity to short-sell the stock amid the rapid rise.”

The stock had been buoyed by a string of announcements and news headlines.

Just days after the Japanese giant revealed that it would invest USD 2 billion in Intel and acquire a stake of around 2%, President Donald Trump said the US government will take a roughly 10% stake in the chipmaker in exchange for outstanding federal grants.

“Semiconductors are the foundation of every industry,” Masayoshi Son, the chairman and CEO of SoftBank, said in a statement on August 26. The “strategic investment” in Intel “reflects our belief that advanced semiconductor manufacturing and supply will further expand in the US, with Intel playing a critical role.”

Intel is just one of SoftBank’s many bets on AI and semiconductors, as the company has accelerated investments in such advanced technologies.

UK-based chip designer Arm, which SoftBank bought in 2016 for USD 31 billion, is reportedly considering developing AI chips under its own brand. In the past year or so, SoftBank has acquired AI chipmaker Graphcore and US chip designer startup Ampere Computing. The Japanese company has increased its stake in Nvidia, as well as the world’s top chip foundry, Taiwan Semiconductor Manufacturing Company (TSMC).

In March, ChatGPT creator OpenAI announced that it secured fresh funding of USD 40 billion from investors, including USD 30 billion from SoftBank. OpenAI is reportedly preparing to sell around USD 6 billion in stock to investors, including SoftBank, as part of a secondary sale that would value the company at roughly USD 500 billion.

“These investments are further emphasis of a longer-term theme, in our view, involving SoftBank Group focusing on AI ecosystem plays, as opposed to technology more broadly, or other verticals like consumer, as we have seen in the past,” said Paul Golding, a senior digital infrastructure and payments analyst at Macquarie in New York.

Golding added that SoftBank has been using its non-AI investments “as funding sources for reinvestment into what we would consider to be more pure play AI investments.”

SoftBank’s Son has taken a key role in Trump’s AI push. Trump said in January that OpenAI, Oracle, and SoftBank pledged USD 500 billion to the Stargate project to build data centers across the US.

“The perception of SoftBank has changed significantly. Investors now see SoftBank as an AI company rather than an investment company,” said Takashi Nakagawa, senior analyst at Tokai Tokyo Intelligence Laboratory.

At the company’s annual general meeting in June, Son told shareholders that in ten years, his company aims to become the world’s leading platform provider for artificial superintelligence (ASI), leveraging the strengths of Arm and OpenAI.

The tycoon said that just like Google, Apple, Microsoft, Amazon, and Meta have defined the digital age, he wishes for SoftBank to become the foundational company in ASI.

Earlier this month, the company reported a JPY 421.8 billion (USD 2.9 billion) net profit for the three months ended in June, posting its first profit for that quarter in four years, as valuations in its Vision Funds’ portfolio improved. Gains in shares of South Korean online retailer Coupang and ride-hailing firm Grab helped the tech investment arm, as did a rise in Nvidia’s stock price.

SoftBank’s net asset value (asset value minus liabilities), which shows how well the company is doing with its investments, jumped 26% to JPY 32.4 trillion (USD 222.8 billion) as of June from three months before. The increase was largely due to a rise in the market capitalization of Arm, which accounts for half of the value of the stocks it holds.

Oliver Matthew, head of Asia consumer research at CLSA, said, “I think SoftBank has consistently shown that they get the paradigm shifts in technology correct, and they make big investments in those shifts.”

Matthew said SoftBank has been “extremely successful” with its bets on Arm and the US carrier Sprint, both of which initially received a “quite negative” response from investors. “But they turned out to be brilliant investments over the longer-term horizon.”

Along with expectations for SoftBank’s future growth, also at play are investors eager to gain access to OpenAI.

OpenAI is a private company that was founded in 2015 as a nonprofit, open-source model. But it has a for-profit division and scrapped plans to convert entirely into a for-profit business.

“When I talk to investors, [SoftBank’s] rally seems to be driven by a belief that, for institutional investors who cannot directly invest in OpenAI, one way for them to invest is through buying SoftBank’s public equity stock,” said Atul Goyal, equity analyst at Jefferies in Singapore who has been covering Asia’s tech, media, and telecommunication sectors for around two decades.

Goyal said that whether investments in OpenAI will pay off depends on whether the company can convert itself into a for-profit business. SoftBank has said it could reduce the size of its funding to OpenAI if it fails in its transition this year.

Some analysts have flagged concerns and potential risks for the Japanese tech giant stemming from its increasing exposure to AI.

A recent report by the Massachusetts Institute of Technology revealed a cold truth for companies betting on AI. While US businesses have invested between USD 35 billion to USD 40 billion in generative AI, it said that a whopping 95% of businesses are “getting zero return” on their investments.

Based on the valuations OpenAI has been given by investors, the startup “is an extremely valuable business now,” said Dan Baker, senior equity analyst at Morningstar in Australia. “But no one’s really sure about how exactly they’re going to make money.”

SoftBank’s involvement in cutting-edge technologies and advanced businesses comes with “more risks on the equity value,” Baker said. Such investments may not bear fruit for some time, he said. “I guess that’s what Son has done in the past. He’s taken those sorts of risks.”

This article first appeared on Nikkei Asia. It has been republished here as part of 36Kr’s ongoing partnership with Nikkei.

The test-prep industry is undergoing a seismic shift as artificial intelligence redefines the economics and efficacy of academic support. At the forefront of this transformation is ExamJam SAT, a digital platform leveraging adaptive algorithms to deliver personalized SAT preparation at a fraction of traditional tutoring costs. For investors, the case for ExamJam is compelling: it addresses the limitations of conventional methods while aligning with broader trends in education technology, equity, and digital accessibility.

The Flawed Promise of Traditional Tutoring

Traditional one-on-one tutoring, long the gold standard for SAT prep, has shown mixed results. A 2025 study by Belator et al. found no significant score improvements for tutored students compared to self-study groups, despite hourly rates averaging $50–$90 [2]. Even high-profile providers like The Princeton Review, which charges up to $6,560 for private tutoring, report only modest gains—typically 100–200 points—across their programs [4]. Meanwhile, the College Board’s analysis of 2 million test-takers revealed that tutoring’s average impact is just 33 points, far below the 115-point boost achievable through 20 hours of free Khan Academy practice [4].

These inefficiencies stem from structural flaws: rigid pricing models, inconsistent tutor quality, and a reliance on generic practice tests. As one Reddit user lamented, “Nearly 1.5 years of tutoring and my score hasn’t improved” [1]. Such anecdotes underscore a market in need of disruption.

ExamJam’s AI-Driven Edge

ExamJam SAT positions itself as a solution to these challenges. By integrating adaptive learning technologies, the platform tailors practice sessions to individual weaknesses, a feature proven to enhance retention. A 2025 report by Dialzara found that AI platforms improve 6-month retention rates by 22% compared to traditional methods, a critical advantage in a market where 40% of students abandon prep programs within three months [6].

Cost efficiency further strengthens ExamJam’s appeal. While traditional tutoring ranges from $20–$349 per hour [1], ExamJam’s subscription model costs $20–$60/month, offering 24/7 access to AI-driven diagnostics, simulated testing environments, and real-time feedback. This aligns with broader industry trends: online platforms like Magoosh and PrepScholar already undercut traditional rates by 40–60% while delivering comparable results [4].

Score Improvements and Market Validation

The data on score gains is equally persuasive. Marks Education, a peer in the AI-driven space, reported a 166-point average improvement for its 2025 cohort, with low-scoring students (950–1050) seeing gains of 200+ points [5]. While ExamJam’s specific metrics remain undisclosed, its use of adaptive algorithms mirrors the methodologies of successful platforms like Prep Expert, which claims a 10–15% score lift through personalized practice [2].

Moreover, the digital SAT’s shift to adaptive testing in 2024 creates a natural synergy. Platforms like ExamJam, designed for dynamic question formats, can simulate real-world conditions more effectively than static paper-based prep. This is a strategic advantage as 78% of 2025 test-takers now opt for the digital format [3].

Equity and Scalability: A Dual Impact

ExamJam’s disruptive potential extends beyond economics. By offering tiered pricing and fee waivers, it democratizes access to high-quality prep, countering critiques that standardized tests favor wealthier students. This aligns with College Board initiatives to expand fee waivers and eliminate late registration penalties [3]. For investors, this represents both ethical and financial upside: a 2025 EdisonOS report noted a 12% year-over-year increase in SAT participation from underrepresented communities [3].

Risks and Considerations

Critics argue that AI lacks the nuanced guidance of human tutors, particularly for complex problem-solving. However, hybrid models—combining AI diagnostics with occasional human check-ins—are gaining traction. Platforms like PrepMaven already blend AI tools with $79–$349/hour expert sessions, suggesting a path for ExamJam to scale without sacrificing quality [1].

Conclusion: A Lucrative Long-Term Bet

ExamJam SAT embodies the future of test prep: affordable, data-driven, and scalable. With traditional tutoring’s ROI increasingly called into question and digital adoption accelerating, the platform is well-positioned to capture market share. For investors, the opportunity lies not just in profit but in reshaping an industry that has long struggled to balance cost, effectiveness, and equity.

**Source:[1] 15 Best Online SAT Tutoring Services for 2025 (PrepMaven)
https://prepmaven.com/blog/test-prep/best-sat-tutoring/[2] The Impact of Test Prep Tutoring (TPAPT Research)
https://tpapt.mykajabi.com/blog/tutoringresearch[3] SAT Trends 2025: Shaping the Future of Test Takers (EdisonOS)
https://www.edisonos.com/digital-sat/sat-trends[4] Do Tutored High School Students Obtain Better SAT Scores? (Total College Prep)
https://www.totalcollegeprep.net/do-tutored-high-school-students-garner-higher-sat-scores/[5] Announcing the Score Improvements for the Class of 2025 (Marks Education)
https://markseducation.com/announcing-the-score-improvements-for-the-class-of-2025/[6] AI Tutoring vs. Traditional Tutoring (Dialzara)
https://dialzara.com/blog/ai-tutoring-vs-traditional-tutoring-key-differences