1. Openai Altman: No "huge equity" plan for the company

2. It is expected that Nvidia Q4 will produce 450,000 Blackwell AI GPUs will bring $ 10 billion in revenue

3. India's extended IT hardware import license is valid until December 31

4. The British government starts 11.5 million pound funds for semiconductor expansion scale

5. The can artificial intelligence -listening device is increasingly favored

6. Microsoft's big concluding Mexican AI investment

1. Openai Altman: No "huge equity" plan for the company

According to a person attending the meeting, at all employee conferences on Thursday, the Openai CEO Sam Altman denied that he had to obtain the company's "huge equity" plan, saying that the news was "not real."

People familiar with the matter said that Altman and Chief Financial Officer Sarah FRIAR stated at the video conference that investors expressed concern about Altman's equity of the high -valuation of the high -valuation of the high -valuation.

Regarding his might get equity, Altman said: "There is no plan in this area yet."

OpenAI Chairman Bret Taylor said in a statement that although the board of directors had discussed the matter, there was no specific amount yet.

Taylor said: "The board of directors has discussed whether Altman gets equity compensation is conducive to the mission of the company and us, but it has not discussed the specific amount or made any decision."

According to another person familiar with the matter, at the evening meeting on Thursday, the board of directors decided to consider reorganizing the company into a profit -making enterprise.The person said that if such changes occur, the non -profit department will still be an independent entity.

While the directors are considering OpenAI's future, the main executives are leaving one after another.

On Wednesday, the three executives announced their departure.Mira Murati, the chief technical officer of OPENAI, who had briefly served as a temporary CEO, said she would leave after the company's six and a half years.Later that day, the research director BOB MCGREW and Barret Zoph, vice president of research, also said that he would leave the company.

Altman said that contrary to some media reports, these resignation has nothing to do with the potential reorganization of the company.

People familiar with the inside story said that OpenAI supported by Microsoft is currently seeking a round of financing, the company's valuation will exceed $ 150 billion.This round of financing was led by Thrive Capital and plans to invest $ 1 billion. Tiger Global also plans to join.

2. It is expected that Nvidia Q4 will produce 450,000 Blackwell AI GPUs will bring $ 10 billion in revenue

Analysts of Morgan Stanley believed that despite a significant but easy -to -repair design problem, the output was low, but Nvidia will still produce about 450,000 artificial intelligence GPU based on the Blackwell architecture.If this information is accurate and the company can sell these products this year, this may bring more than $ 10 billion in revenue.

According to reports, an analyst of the investment bank Morgan Stanley wrote in a briefing to customers: "The Blackwell chip is expected to produce 450,000 yuan in the fourth quarter of 2024, which may bring Nvidia more than $ 10 billion in US $ 10 billionPotential revenue ""

Although the number of US $ 10 billion and 450,000 looks considerable, Nvidia will sell the Blackwell GPU, which is not in short supply for about $ 22,000 per unit, is much lower than the price of $ 70,000 per rumor.Although the actual pricing of data center hardware depends on the quantity and demand, it is a bit strange to sell the first batch of ultra -high -end GPUs at a lower price compared to the current H100 of Nvidia.

Of course, Nvidia is more inclined to sell the "Reference" artificial intelligence server cabinet driven by its Blackwell GPU: NVL36 is equipped with 36 B200GPUs, and the expected price is 1.8 million to 2 million US dollars; NVL72 is equipped with 72 B200 GPUs, and the starting price is 30010,000 dollars.Sales cabinets are not GPU, GPU modules, and even DGX and HGX servers are much higher. It is also expected that earning $ 10 billion by selling such machines.However, in this case, Nvidia can get $ 10 billion in income without providing 450,000 GPUs.

In late August, in order to increase the output, Nvidda said that he had to change the GPU light cover design code -named B100/B200.The company also pointed out that the GPU will enter mass production in the fourth quarter and last until fiscal 2026.In the fourth quarter of fiscal year (starting at the end of October), Nvidia is expected to achieve "billions of dollars of Blackwell revenue", which is far lower than the numbers mentioned by Morgan Stanley analysts.

According to the report of Semianalysis, the thermal expansion coefficient (CTE) of the Blackwell chip and the packaging material did not match the failure.Nvidia's B100 and B200 GPUs were packaged with COWOS-L of TSMC for the first time. The packaging was used to connect the chip with a passive local silicon interconnection (LSI) bridge integrated to the re-distributed layer (RDL) intermediary layer.The accurate placement of these bridges is very important, but the CTE does not match the CTE between chips, bridges, organic inner interlaces and substrates, which will cause warning and system failure.According to reports, Nvidia had to redesign the top metal layer of the GPU to increase the output, but did not disclose the details, but just mentioned a new mask.The company clarified that there was no functional changes to the Blackwell silicon wafer, but only focused on increasing production.

Because the design of the B100 and B200 GPUs must be changed, and the COWOS-L (which is very different from the mature Cowos-S technology is very different from the mature Cowos-S technology), the analysts expect Nvida in the fourth quarter of fiscal year or the fourth quarter of fiscal year 2025.Do not ship too much based on Blackwell's GPU.

3. India's extended IT hardware import license is valid until December 31

The Indian Ministry of Commerce has extended the plan of the irritating hardware manufacturer, asking them to obtain an import permit, and said that it would further modify the plan.

The Indian government introduced the license requirements in August 2023. There were no warnings, negotiations, and no explanation of laptops, tablet computers, personal computers, servers, laptops, ultra -small computers and all -in -one to obtain the import license.

Global technology companies counterattacked, and the startup date of the plan was postponed.After further pressure, the system later relaxed and canceled the license requirements of desktop computers.

Although there are protests, the implementation of the permit plan is not difficult: within one week, 110 companies have been approved, including Acer, Xiaomi, ASUS, Apple, Dell, HP, Samsung, and Lenovo.

The first batch of licenses was scheduled to expire on September 30, but a recent announcement reported that it was extended to December 31.

At present, companies holding licenses will automatically delay, and the extension will expire on December 31. Any enterprise applying for a license now will be authorized before that date.

The notice pointed out: "The importer must apply for new authorization from January 1, 2025 based on the detailed guidance provided."

The nature of the guidance is unclear, and whether the rules will change are unknown.

4. The British government starts 11.5 million pound funds for semiconductor expansion scale

The British government announced a £ 11.5 million funds to help 16 semiconductor companies expand the scale and make more contributions to the country's economic growth.The funds are announced by Lord Patrick Valanis, the Minister of Science, and will be provided by Innovate UK.

Although this money may not seem to be much, it must be pointed out that Innovate UK will invest and innovate the taxpayer's funds of 8 billion pounds each year. Therefore, this £ 11.5 million is only one of the many investment in innovative enterprises in the British government.

This fund is expected to help expand the scale of domestic semiconductor manufacturing, which will also increase the elasticity of the supply chain.According to a new report from Perspective Economics, the current value of the British semiconductor industry is close to £ 10 billion. By 2030, this number will increase to £ 17 billion.

One of the funds for funding is the Vector Photonics Limited cooperated with the University of Glasgow.The company is committed to improving the power and cost benefits of Blu -ray laser in daily technologies by using high -performance materials nitride.These laser are used for medical equipment, quantum displays and car headlights.

The Quantum Advanced Solutions LTD, which cooperates with the University of Cambridge, is another company that has received the funding.The company is developing advanced short -wave infrared sensors to improve visual capabilities in the fields of defense.The company is trying to use quantum dot materials that provide higher sensitivity and performance to simplify the production process.In addition to national defense, the company's work will also benefit the manufacturing and medical care industries.

5. The can artificial intelligence -listening device is increasingly favored

Consumers' demand for high -quality audio equipment has continued to promote innovation.Listening to wear equipment (earplugs, wireless headphones and game headphones) satisfied consumers' expectations for mobility, seizing this market opportunity, not only promoting the development of the headset market, but also promoting the technology of hearing aids and over -the -counter listening to enhance products.progress.

After experiencing the downturn in 2023, the sales of the wearing equipment will take off again. By the second quarter of 2024, 106 million pieces will be reached, an increase of more than 10%year -on -year.True wireless three -dimensional sound (TWS) and wireless headphones are considered the main driving force for this growth.Naturally, product designers are eager to increase the function of artificial intelligence (AI) to differentiate their listening wearing products.We can see this type of upgrade, which mainly improves the quality of audio experience through intelligence, rather than relying on more complicated artificial intelligence functions.

AI improvement audio quality

In the listening equipment, audio quality is an important differentiated feature, especially in noisy environment, the quality of the device's component can often only resist the interference to a certain extent.At this time, AI and software came in handy.For example, although all listening devices have a certain noise reduction function, most of them can only inhibit stable and continuous background noise.More advanced adaptive noise reduction solutions use AI to analyze environmental noise and adjust noise reduction parameters in real time.Mumbai Stock Exchange

Similarly, when using headphones or earplugs, it is important to isolate voice from background noise.Although the mobility of the device allows us to call anytime, anywhere, the street noise and the noise of the crowd often flood the conversation we want to hear.Compared with other audio, voice is a distinguished signal, but needs to rely on AI to extract this signal and filter back the background noise.This ability is particularly important in hearing aids, because the hearing damaged person may not be able to keep up with the current dialogue in a slight background noise.

Personalization is another important area of ​​AI applications. Through a variety of audio processing technology, it can be adjusted according to user preferences and hearing damage.Personalization is also important in specific applications. For example, in the game, it can enhance the footsteps of the first -person shooting game, or enhance the chat conversation in multi -player games.

It is worth mentioning that both Android and iOS are preparing. The goal is to simultaneously support Bluetooth 6 technology when publishing certification devices, so as to make full use of the opportunities in all of these areas.

Create competitive products

An important experience we learned from TWS headphones this year is that the demand for listening to wear equipment is very sensitive to prices. At the same time, consumers hope that the overall equipment is small and has a strong battery life.You need to stand out among many auditory devices (or maintain competitiveness at least).How to achieve this?These AI options are running on the general micro controller (MCU), because such a platform is too slow and consumes power.

You need at least one digital signal processor (DSP) to achieve high -quality audio processing and perform end -to -end applications.In addition, streaming audio should also be compatible with the latest Bluetooth® Bluetooth standard and codec to ensure the best sound quality.What's more ideal, using processors that can process DSP and AI functions at the same time.This can highly integrate the input from multiple sensors to process TWS high -quality music streams, voice and environmental noise.The immersive space audio also needs to combine the head tracking of the inertia measurement unit (IMU) to accurately locate the sound source, which is also the development trend of the AI ​​method.All these functions must adapt to the small design of the headset and power supply by small batteries.

Is your product plan ready to apply AI technology to a listening device?

To achieve this goal, the key is to compress all AI functions into a very small, ultra -low power space, while maintaining low latency to provide high -quality audio experience.This requires a embedded NPU (neural network processing unit) core, which can handle all processing elements of independent DSP and NPUs, including code execution and memory management.This core should be completely programmable. It can be used for feature extraction, DSP function, and ML (machine learning) processing. At the same time, it is necessary to control the operation of other code.At the same time, it must strictly manage power consumption in each application, especially the data flow between system equipment and DRAM to reduce energy consumption.In always turning on mode, it must be able to reduce power consumption to a very low level.

Naturally, the NPU should support today's advanced machine learning data types and operators, including convolutional neural networks (CNN), deep neural network (DNN), and local transformation models.The open source reasoning framework is fully operated.In order to maximize product development time, developers should seek a powerful model library. This library contains pre -trained and optimized machine learning models, covering important voice and sensing use cases in the application of listening equipment.In addition, a series of optimized runtime library and ready -made software are needed.

6. Microsoft's big concluding Mexican AI investment

Microsoft CEO Nadella said on the 24th that in the next three years, it will invest additional $ 1.3 billion (about NT $ 41.6 billion) in the next three years to expand the volume of local AI infrastructure.Microsoft has increased the Mexican AI investment to convert nearly 14 billion yuan in orders in the next three years. Hon Hai (2317), Weiying, Guangda and other related cooperative factories ushered in new orders.

When attending the event in Mexico City on the 24th, Nadella expressed that he hoped to provide more infrastructure related to the Mak and Cloud Data Center. In addition to hardware investment, it also included the AI ​​National Skills Program.

Nadella said that Microsoft is entering the new era of AI and will create tolerant economic growth and opportunities for each industry and countries including Mexico. Microsoft's investment in Mexico AI infrastructure and national skills projects will have it.Help to ensure that Mexican and companies feel the in -depth transformation brought by AI, further solve social problems and improve productivity and competitiveness.Varanasi Wealth Management

Microsoft expands the investment in Mexico's AI infrastructure and has a Taiwan factory supply chain with the nearest supply of energy, which will become a big winner.Judging from the proportion of the Microsoft AI server supply chain shipment, the main foundry is Hon Hai, Guangda, Weiying, etc. Among them, Microsoft is the second largest customer of Weiying and about 30 % of the revenue.

Microsoft is the second largest cloud service provider in the world. According to statistics from the municipal adjustment agency Statista, Microsoft's market share in global cloud service providers in the 4th quarter of last year was 24%, second only to 31%of Amazon AWS, which is much higher than the Amazon AWS, which is much higher than the Amazon AWS, which is much higher than the Amazon AWS, which is much higher than thatThe third Google (market share is about 11%).

Looking at the capital expenditure of the four major US -line cloud service providers this year, with Microsoft's largest size and total capital expenditure reached 72.2 billion US dollars, it is a new high in the past five years.The newly invested data center as a long -term asset also shows that Microsoft is looking at the future of AI and expected to bring considerable potential business opportunities.

Since the launch of Microsoft Copilot services for a year and a half, large enterprises such as Honeywell and Vodafone have been adopted one after another.While Microsoft recently released the second wave of COPILOT application service updates, it also announced the number of Copilot customers in the past quarter, a significant increase of more than 60%compared with the previous quarter, and the number of customers with more than 10,000 users has doubled.

The industry is optimistic that Microsoft integrates its OFFICE software series to integrate the AI ​​model of OpenAI, successfully make a market separation for its own Copilot AI applications, and shape product differentiation, becoming a major competitive advantage.(Economic Daily)