Venture Bytes #116: US Start-ups May Have an Abundance Problem

This maxim by pioneering physicist Ernest Rutherford could just as well describe the ingenuity of DeepSeek, a Chinese AI start-up that turned necessity into an asset. Employing a MacGyver-like approach, DeepSeek built a highly capable AI reasoning model despite severe resource constraints—leveraging efficiency and architectural innovation instead of brute-force compute. This raises a critical question: Are US VC- backed startups suffering from an abundance problem?

The core argument isn’t that abundant capital is inherently negative. Instead, the paradox lies in the diminishing returns of excessive funding. When money is no longer a limiting factor, the urgency to innovate from first principles wanes. Historical evidence and current trends indicate that constraints — whether in funding, hardware, or regulatory environments — can compel firms to focus on fundamental efficiencies and robust problem-solving. In a sense, scarcity can be a catalyst for true innovation, pushing teams to explore alternative solutions and optimize operations in ways that excess capital might inadvertently discourage.

China has a long history of turning constraints into competitive advantages. Huawei’s success in the 5G arena provides another compelling example of Chinese innovation despite geopolitical challenges and restrictions. According to a report on 5G Patent Race by New York-based data analytics company LexisNexis, Huawei ranked first among the top 50 5G ultimate patent owner and standards contributing entities. By 2023, Huawei had launched 300+ commercial 5G networks worldwide, serving over 1.6 billion users. In 2024, Huawei surpassed Ericsson and Nokia in the consumer business game with the highest progress rate.

BYD in electric vehicles (EVs) is another example. With limited early funding compared to Tesla, BYD focused on battery innovation and cost efficiency. Despite US trade restrictions on Chinese-made connected vehicle software, BYD has managed to thrive, with innovations like its Blade Battery—a revolutionary design that improves heat dissipation—and “God’s Eye,” a next-generation driver-assistance system.

The biotech sector is also witnessing what could be called a “DeepSeek Moment.” The Wall Street Journal recently highlighted how Summit Therapeutics’ cancer drug licensed from China’s Akeso, outperformed Merck’s $30 billion blockbuster Keytruda. More importantly, this is not an anomaly— Chinese pharmaceutical firms now account for   of major pharma licensing deals, up from just 5% in 2020. Bloomberg Economics research shows that China has overtaken the US in international patent applications, which are stretching across a broader range of areas. Of 13 key technologies tracked by Bloomberg researchers, China has achieved a global leadership position in five of them and is catching up fast in seven others.

In the US, venture capital rounds have grown significantly in size. Data from PitchBook indicates that the median Series D and beyond rounds in the US have surged from just more than $20 million in 2016 to roughly $97 million in 2024 — a nearly 5-fold surge.

The trend of hyper-funded startups reflects a broader shift in VC toward “moon-shot” bets. For example, Safe Superintelligence raised a $1 billion Series A round in September 2024, when the company was only 3 months old, and is reportedly exploring a new round at a $20 billion valuation. California-based Xaira Therapeutics also raised a $1 billion Series A round within the first year. While these companies are undoubtedly tackling ambitious problems and developing cutting-edge technologies, the scale of their funding raises important questions about the implications of such rapid capital infusion.

A 2012 study analyzing science funding efficiency found that measurable outcomes, such as advancements in life expectancy or weather prediction accuracy, scale sub-linearly with resource allocation.  For instance,  despite  exponential  increases  in healthcare R&D spending over 170 years, human life expectancy improved linearly rather than exponentially. Similarly, investments in numerical weather prediction systems over 50 years achieved incremental reliability gains that did not proportionally match funding growth 16. This principle extends to technology-driven ventures. While initial capital enables critical infrastructure (e.g., GPU clusters for AI training), overfunding can lead to “resource bloat,” where marginal technical improvements fail to justify escalating costs.

AI’s Next Move is Local

The AI revolution is expanding beyond the cloud and large servers, and into billions of smaller devices at the dynamic edge. With the number of connected IoT devices expected to reach 40 billion by 2030 from 16 billion in 2023, per IoT Analytics, this shift represents a big opportunity for start-ups and investors.

The next wave of Industry 4.0 demands speed, efficiency, and reliability—precisely what edge AI delivers. Unlike traditional cloud computing, which centralizes processing, edge AI brings intelligence directly to the data source — whether in autonomous vehicles, industrial automation, or smart devices. By handling computations on-site, it reduces latency, enhances real-time decision-making, and cuts reliance on cloud infrastructure.

Manufacturing, healthcare, automotive, and retail sectors are leading the way with edge deployments. Manufacturing accounts for around 31% of Edge AI market revenue, making it the largest adopter of the technology, according to the 2024 State of Edge AI Report by the Dutch knowledge platform Wevolver. Healthcare relies on edge AI for life-critical applications, such as real-time patient  monitoring  and  robot-assisted  surgeries,  ensuring HIPAA-compliant data processing and faster response times. In manufacturing, edge computing supports predictive maintenance through tinyML, reducing downtime and operational costs.

Edge AI solves a key challenge in the automotive sector. Connected cars generate an overwhelming 1–2 terabytes of data per day—far exceeding current cellular data limits. Uploading everything to the cloud is neither practical nor cost-effective. The real challenge lies in filtering and prioritizing the most valuable data for performance and reliability improvements. Edge AI solves this by processing data locally, sending only critical insights to the cloud. With edge computing permeating a lot of industries and application areas, the global market for the technology is expected to reach $269.8 billion by 2032 from $27 billion in 2024, growing at a CAGR of 33.3%, per Fortune Business Insights.

Edge AI is also set to be a key catalyst for the smartphone industry. The smartphone industry saw a 4% decline in global shipments in 2023, following a 10% drop in 2022, according to Canalys and IDTechX. In response, leading technology companies are already incorporating AI into their devices. Samsung unveiled the Galaxy S24 series with AI capabilities in January 2024, while Apple’s iPhone 16, launched in September 2024, introduced the A18 chip, which provides up to 30% more processing power for AI tasks compared to the A17 chip in the iPhone 15.

Smartphone sell-through grew by 4% YoY in 2024, recovering from the weakest year in a decade in 2023. While the limited presence of AI-powered smartphones in 2024 did not fully drive this growth, its full impact will be felt in the coming years. Such smartphones are currently restricted to the premium segment; however, their adoption will extend to mid-range devices, driving substantial YoY sales growth as consumers shift from non-AI to AI-enabled smartphones. Semron AI, a German startup, offers edge AI chips for smartphones, VR sets, and earbuds. Semron’s technology promises up to a 20x increase in chip efficiency.

Various start-ups are well-poised to capitalize on this opportunity. California-based EnCharge claims its AI accelerators use 20 times less energy to run workloads compared with other chips on the market. The company, which recently raised a $100 billion Series B round, expects to have the first of those chips on the market later in 2025.

South Korean startup DeepX, with partnerships including Hyundai Kia Motors and Posco DX, achieves over 10 TOPS/W computational efficiency with under 3W power consumption, delivering superior AI recognition accuracy. Israel-based Hailo, a key player in edge AI, serves industries like automotive, security, industrial automation, and smart retail. The company is also expanding into consumer electronics through collaborations such as Raspberry Pi. As of 2024, Hailo has built a strong customer base exceeding 300 companies.