Nvidia continues to dominate AI investment talk, with analysts highlighting its dual role in artificial intelligence and quantum computing. The company does not build its own quantum hardware but supports the field through products like NVQLink and the CUDA-Q software platform. Its Ising model reportedly speeds up quantum processor calibration and performs error correction up to 2.5 times faster than traditional methods. Management expects revenue growth of 70% in the next fiscal year, and the stock trades at 24.4 times forward earnings.
SpaceX is expanding into AI infrastructure in a big way, building large data centers powered by Nvidia chips. The company aims for over 2 gigawatts of compute capacity by year end, growing from $2.6 billion in AI revenue in Q2 2026. Partners including Google and Anthropic are helping fill the new computing capacity, which is backed by $85.7 billion in IPO proceeds and $25 billion in notes.
On the quantum side, Diraq is taking a different approach, believing quantum computers can reduce AI's energy demands. Its current system runs on just eight qubits inside standard data centers with under 20 kW of power use. The company uses silicon chips built on standard CMOS foundry lines and is integrating its processor with a Dell high-performance computing cluster, targeting millions of qubits at under a dollar each.
Microsoft faces a practical challenge as its 365 Copilot tools expose data risks tied to weak permissions. Many organizations have years of overshared files and outdated access rules that were never cleaned up. Security expert Saviona Pereira notes the issue is usually not carelessness but a lack of strong data governance. Without fixing these foundations, widespread deployment of AI assistants remains risky.
In model development, Supersonic Labs launched Julia 1, a 144.3M-parameter open decision model that runs entirely on a CPU. It handles three tasks through one API: choosing from options, scoring on an ordered scale, and answering yes or no. Training cost roughly $104 in cloud GPUs, and the model is licensed under Apache 2.0 with a $0.025 per million tokens API planned.
Liquid AI is focusing on speed, releasing LFM2.5-VL-3B-DSpark, a draft model that uses speculative decoding to accelerate its vision-language model. The drafter adds about 280M parameters and achieved up to 3.13x faster decoding on Apple silicon and up to 2.66x on an NVIDIA H100, with output quality unchanged under greedy decoding.
On the commercial front, GoKwik's AI Calling feature on WhatsApp helped 30 brands double their abandoned-cart recovery rates. The platform handles over one million AI-powered calls per month, and customers who received both AI calls and WhatsApp messages were 2.5 times more likely to buy.
Key Takeaways
- Nvidia expects 70% revenue growth in its next fiscal year, driven by strong AI demand, and trades at 24.4 times forward earnings.
- SpaceX is building AI data centers with Nvidia chips, targeting over 2 gigawatts of compute by year end, with Google and Anthropic as partners.
- Diraq's eight-qubit quantum system runs under 20 kW in standard data centers and aims for millions of qubits at under a dollar each.
- Microsoft 365 Copilot can expose sensitive data when permissions are poorly managed, requiring stronger data governance before wide adoption.
- Supersonic Labs' Julia 1 is a 144.3M-parameter decision model running on CPUs, trained for about $104 and licensed under Apache 2.0.
- Liquid AI's DSpark draft model speeds up LFM2.5-VL-3B by up to 3.13x on Apple silicon and 2.66x on NVIDIA H100.
- GoKwik's AI WhatsApp calls help 30 brands double abandoned-cart recovery, processing over one million calls monthly.
- Nvidia's Ising model performs quantum error correction up to 2.5 times faster than traditional methods.
- SpaceX's AI revenue reached $2.6 billion in Q2 2026, backed by $85.7 billion in IPO proceeds and $25 billion in notes.
- Rollins stock fell 6.4% to $30.40 on September 24, 2026, down 48% year-to-date as AI threatens traditional pest control models.
Nvidia leads AI and quantum computing investment
Nvidia is not building its own quantum computing unit but supports the field with products like NVQLink and CUDA-Q software. Its Ising model speeds up quantum processor calibration and performs error correction up to 2.5 times faster than traditional methods. Management expects revenue to grow 70% in the next fiscal year due to strong AI demand. The stock trades at 24.4 times forward earnings.
Nvidia stock seen as top AI and quantum pick
Nvidia stands out as a rare company involved in both artificial intelligence and quantum computing. It supports quantum research through products like NVQLink and the CUDA-Q software platform. The company expects revenue growth of 70% in its upcoming fiscal year. Nvidia trades at 24.4 times forward earnings, which analysts find attractive.
Rollins stock drops as AI threatens pest control
Rollins, a major U.S. pest control company, saw its stock fall 6.4% to $30.40 on September 24, 2026. The stock is down about 48% this year and 53% from its high of $66. Piper Sandler downgraded the rating to Neutral due to valuation concerns. The CEO remains optimistic about the brand and new business pipeline despite debt and market share worries.
SpaceX expands AI data center with Nvidia chips
SpaceX is building large AI data centers powered by Nvidia chips for AI workloads. The company targets over 2 gigawatts of compute by year end, growing from $2.6 billion in AI revenue in Q2 2026. SpaceX is backed by $85.7 billion in IPO proceeds and $25 billion in notes. Partners like Google and Anthropic are helping fill the new computing capacity.
Diraq says quantum computers may reduce AI energy use
Diraq believes quantum computers can help cut AI's energy use by improving training data creation. Its current system has eight qubits and runs in standard data centers with under 20 kW power use. The company uses silicon chips made with standard CMOS foundry lines, aiming for millions of qubits at under a dollar each. Diraq is integrating its processor with a Dell high-performance computing cluster.
Microsoft 365 AI tools expose data risks from weak permissions
Microsoft 365 Copilot can find and share sensitive documents, emails and chats if permissions are not set properly. Many companies have years of overshared files and old access rules that were never cleaned up. Security expert Saviona Pereira says the problem is not usually carelessness but a lack of strong data governance. Security teams and business units often do not communicate well, leaving unclear ownership of data. Fixing these foundations is essential before AI assistants are deployed widely.
Supersonic Labs releases Julia 1 decision model for CPUs
Supersonic Labs launched Julia 1, a 144.3M-parameter open decision model that runs on a CPU. It handles three tasks through one API: choosing from options, scoring on an ordered scale, and answering yes or no. Training cost about US$104 in cloud GPUs. In tests, it beat reference scores on AG News and DAIR Emotion but trailed badly on Banking77. The model is licensed under Apache 2.0 and a $0.025 per million tokens API is planned.
Three Japanese AI healthcare stocks to consider
With Japanese bond yields at a 30-year high, investors are seeking growth themes with real-world use cases. Chugai Pharmaceutical, Asteria Biotech and Fujifilm Holdings are three Japanese companies using AI in healthcare. All three focus on AI-assisted diagnosis, treatment planning and patient engagement. These stocks represent a small sample of AI healthcare opportunities in Japan. The article does not give investment advice but encourages further evaluation.
Liquid AI speeds up vision-language models with DSpark
Liquid AI released LFM2.5-VL-3B-DSpark, a draft model that uses speculative decoding to speed up its LFM2.5-VL-3B vision-language model. The drafter adds about 280M parameters and checks proposed tokens in one pass. It achieved up to 3.13x faster decoding on Apple silicon and up to 2.66x on an NVIDIA H100. Output quality stays the same under greedy decoding. The drafter uses a simplified attention-only architecture with 4 layers and a block size of 8 or 9.
AI WhatsApp calls help e-commerce recover lost sales
GoKwik's AI Calling feature on WhatsApp helped 30 brands double their abandoned-cart recovery rates. The platform handles over one million AI-powered calls per month. Customers who received both AI calls and WhatsApp messages were 2.5 times more likely to buy. The combined approach delivered more than 10 times the incremental return on spend compared with WhatsApp alone. Calls run between 10 am and 9 pm with consent and opt-out controls.
Sources
- This Stock May Be the Best Artificial Intelligence (AI) and Quantum Computing Investment
- This Stock May Be the Best Artificial Intelligence (AI) and Quantum Computing Investment
- Can AI Really Disrupt Pest Control? Wall Street Thinks So, and Rollins (ROL) Just Hit New Lows
- Space Exploration Technologies (SPCX) Expands Its AI Data Center Buildout
- Diraq Says Quantum Computers Could Cut AI’s Energy Use For Data Training
- Securing the data behind the AI: Information protection and insider risk in the Microsoft 365 era
- Supersonic Labs Releases Julia 1: A 144.3M-Parameter Open Decision Model That Runs on a CPU
- Chugai Pharmaceutical And 2 Japanese AI Healthcare Stocks To Own
- Liquid AI Releases LFM2.5-VL-3B-DSpark: Speculative Decoding for Vision-Language Models With Up to 3.13x Faster Decoding
- AI-powered WhatsApp calls boost e-commerce sales
Comments
Please log in to post a comment.