
Crypto News & Market Updates
Today (11/18/2026)
•
London Disseqt raises $8 million to develop AI intelligent entity governance
Deep Tide TechFlow reports that on November 18, according to Sifted, London startup Disseqt completed an $8 million funding round led by Gresham House Ventures. Sifted's headline states that the company is backed by industry insiders from OpenAI and Meta.
Disseqt positions itself as an "AI Assurance Layer," situated between enterprise intelligent agents and risk control functions, covering both pre-deployment and post-launch phases: conducting structured red team testing and pre-production validation before deployment, real-time monitoring and drift detection during operation, and mapping validation results into executable, auditable compliance obligations. The company claims its platform can intercept over 97% of deployment issues before production, improve testing efficiency by more than 40%, with clients mainly in regulated industries.
Due to character-level obfuscation applied by Sifted in the main text for non-subscribers, verifiable information about co-investors and the founding team background is unavailable, so no speculation is made here. As AI intelligent agents are deployed in financial, medical, and other scenarios, "who verifies that they operate according to rules" is becoming an independent category, and Disseqt is betting on this gap in auditing and governance.
•
Paytaca was attacked by a vulnerability generating fake orders, private keys were not leaked, reminder not to engage in suspicious transactions
Bitcoin Cash (BCH) ecosystem self-custody wallet Paytaca issued a security notice stating that on November 17, attackers exploited a P2P Ramp backend vulnerability to generate fake orders; on November 18, another backend vulnerability was exploited, resulting in unsolicited payment requests for users. Paytaca stated that the attacks targeted backend infrastructure and did not directly involve private keys controlling user funds. User private keys are only stored locally on devices and are not stored on backend servers. The team is investigating and fixing the related vulnerabilities and reminds users not to interact with suspicious P2P orders or payment requests, nor to sign transactions not initiated by themselves.
•
Ethereum developers submit EIP draft proposing native support for post-quantum signatures leanSPHINCS in frame transactions
Foresight News reports that Ethereum developer ilitteri has submitted an EIP draft proposing the introduction of the hash-based post-quantum signature scheme leanSPHINCS for frame transactions defined in EIP-8141. According to the draft design, transactions declare the required signature through a dedicated verification frame; the signature data is separated from the transaction body but propagates together with the transaction in the mempool and is included in the block. Each node verifies the signature before executing the transaction. This proposal is still in draft stage, has not been assigned an official EIP number, and the complete technical specification is yet to be supplemented.
•
Antigravity 1st Anniversary Giveaway: All model quotas reset, Carbon initial response speed increased by 1.5 times
Breaking AI News: Varun Mohan, head of Google's Antigravity, announced that to celebrate the one-year anniversary of Antigravity's launch, Google will reset the usage quotas for all models for all users, allowing those who had previously exhausted their quotas to regain available allocations.
This is already the third time this month that Google has collectively reset quotas for Antigravity users. On November 7, Google reset the quotas for Gemini and Argon; on November 14, it reset the Gemini quotas for all users. This time, the reset extends to all models.
Varun also announced that the initial character response speed of Gemini 4 Carbon in Antigravity has increased to 1.5 times the original speed, meaning users' waiting time for the model to start outputting has been reduced by about 33%. Argon will soon receive similar optimizations.
Antigravity was released together with Gemini 3 on November 18, 2025, and today marks its one-year anniversary.
•
Nomura Analysis: Overseas High-End Computing Power Tightens, Domestic AI Chips Welcome Replacement Window
Summary: Overseas computing power tightens, who among Tencent, Alibaba, and ByteDance can fill the high-end GPU gap
TL;DR
· Some overseas data centers' remote computing power supply is now subject to licensing requirements, further tightening domestic companies' access to advanced GPU computing power.
· Nomura warns that if advanced GPU shortages worsen, domestic AI cloud revenue growth in 2027 may face downward risks.
· Adoption of domestic AI chips is expected to accelerate. Nomura cites IDC data showing that in the first half of 2026, Alibaba's Pingtouge held a 6.6% share, ranking third in the domestic related chip market.
· Alibaba and ByteDance's existing GPU resources and financial strength can provide a buffer for model development. Nomura is more concerned about whether Tencent can continue to expand its advanced GPU capacity.
· Independent AI labs may face increased difficulty in obtaining the training computing power needed for next-generation models, with Nomura viewing DeepSeek as a possible exception.
Overseas computing power access tightens, AI cloud growth faces downward risk
On November 16, the U.S. announced interim final rules involving advanced computing chips, including licensing requirements for some cases where advanced computing power is remotely provided through overseas data centers. Even if chips are deployed abroad, if the computing power provided by the data center to specific companies exceeds the stipulated threshold, export or transfer of related chips and servers may require a license.
Nomura's report released on November 17 points out that domestic companies may face increased difficulty in obtaining NVIDIA's advanced GPU computing power through overseas data centers. Alibaba, Tencent, and ByteDance are all increasing investments in AI infrastructure and foundational models. If advanced GPU supply tightens further, model training may be restricted, and the computing power cloud providers can deliver to customers may also be affected.
Nomura expects that if advanced GPU shortages worsen, domestic AI cloud revenue growth in 2027 will face downward risks. The report does not specify the extent of revenue growth adjustment. Nomura stated it will monitor Alibaba and Tencent's weekly earnings communications to assess the impact of changes in computing power supply on their cloud businesses.
Domestic chip adoption may accelerate, Alibaba already has a self-developed foundation
The increased difficulty in obtaining advanced GPUs may prompt companies to accelerate adoption of domestic AI chips. Nomura believes platforms with strong computing power acquisition capabilities and self-developed chip capabilities have more options when adjusting supply arrangements. Alibaba's Pingtouge is a key example mentioned in the report.
Nomura cites IDC data showing that in the first half of 2026, NVIDIA held a 33.8% share, ranking first in the domestic related chip market, Huawei followed closely with 28.5%, and Alibaba's Pingtouge ranked third with 6.6%. Cambricon and Baidu Kunlun chips held shares of 4.6% and 4.2%, respectively.
Caption: Domestic AI chip market share in the first half of 2026. Source: IDC, Nomura Research.
For Alibaba, Pingtouge provides a self-developed chip computing power source, potentially increasing options for subsequent infrastructure deployment. Nomura expects domestic chip adoption to accelerate but does not quantify how much it can fill the advanced GPU supply gap.
Alibaba and ByteDance have computing power reserves, Tencent faces expansion pressure
Changes in computing power supply may also alter the competitive landscape of the foundational model market. Nomura believes independent AI labs may face greater pressure in obtaining advanced GPUs needed for next-generation model training. The report views DeepSeek as a possible exception but does not elaborate further.
Alibaba and ByteDance's existing GPU resources and financial strength may provide some buffer for subsequent model development. Both companies are also affected by tightened overseas computing power access, but existing resources help maintain development pace when new supply is limited.
For Tencent, Nomura is more concerned about whether advanced GPU reserves can continue to expand. If new computing power deployment slows, Tencent may find it harder to maintain its current AI infrastructure investment pace, which could affect model iteration and the speed of new application launches. Going forward, key points to watch in Alibaba and Tencent's earnings communications include whether new computing power arrives as planned and whether the companies will adjust growth expectations for their cloud businesses.
•
Thailand's Meta Corp. Completes Capital Increase
11:34 PM EST, 11/17/2026 (MT Newswires) -- Meta Corp. (SET:META) completed the registration of a 152.74 billion baht increase in its registered capital with Thailand's Department of Business Development on Nov. 16, according to a Tuesday bourse filing.
The solar power company's registered capital rose from 3.07 billion baht to 155.81 billion baht through the issuance of up to 3.68 billion new ordinary shares.
The new shares are allocated for a rights offering of up to 2.44 billion shares, the exercise of up to 739.6 million warrants and the conversion of up to 500 million shares for convertible debentures.
•
Leaked material: Ascend 950DT inference performance reaches 86% of H200
ChainCatcher news: On November 18, AI content account Max For AI reported that a document titled "Ascend 950DT Performance Test and Operational Efficiency Estimation Report" circulated in computing power-related communities. The document is dated October this year and shows it is an internal test report from China Mobile. According to the document, on October 10, a joint team from multiple China Mobile departments visited Huawei's Lian Qiuhu R&D center, using a 64-card Ascend 950DT super node to conduct real tests on Zhipu GLM 5.2 and DeepSeek V4 Pro.
The performance data provided in the document is based on the GLM 5.2, 16K input / 2K output scenario. For single-card total throughput, the Ascend 910B4 achieved 169 tokens/s, 910C achieved 483 tokens/s, and 950DT achieved 1951 tokens/s, approximately 11.5 times, 4 times, respectively, that of the 910B4 and 910C. The document cites vLLM data estimating Nvidia H200 at about 2200 tokens/s under the same scenario, corresponding to about 89% of H200's performance for the 950DT; another set of 7K/2K data is about 82%. The document's final conclusion states that the 950DT's inference performance reaches 86% of the H200.
Regarding cost estimation, the document lists the prices of 8-card servers at bulk procurement prices: 950DT about ¥2.97 million, 910C about ¥2.15 million, and 910B4 about ¥650,000. Including equipment, maintenance, supporting facilities, and IDC costs amortized over five years, the monthly costs are approximately ¥63,600, ¥48,700, and ¥16,100 respectively. Based on this, the document calculates that the unit cost inference performance of the 950DT is about 3.1 times that of the 910C and 2.9 times that of the 910B4. The bare metal rental estimate is ¥84,300 per machine per month for the 950DT and about ¥130,000 for the H200, stating that the 950DT's inference cost-performance ratio is about 30% higher than the H200's.
•
XRP whales transferred about 281 million tokens twice, and the receiving wallets were not marked as exchanges
According to TechFlow, on November 18, CoinDesk reported that an XRP holder transferred approximately 140.35 million XRP in two transactions to two separate wallets, totaling about 281 million XRP, valued at approximately $270 million. Neither of the two receiving wallets has been identified as exchange deposit addresses.
The latest transfer occurred at 23:22 Beijing time on November 17, valued at about $136 million; the other transfer of the same amount took place on November 6. Both wallets were activated by the sending address on October 5, followed by small test transfers. CoinDesk's analysis suggests that these actions are consistent with a pre-planned custody arrangement.
•
King Gen Raises THB270 Million from Share Offering
11:22 PM EST, 11/17/2026 (MT Newswires) -- King Gen (SET:KGEN) has completed a private placement of 200 million ordinary shares, raising 270 million baht, according to a Tuesday filing with the Thai bourse.
The electric vehicle maker allotted 185.2 million shares to Rayong Wire Industries (SET:RWI) and 14.8 million shares to shareholder Jindasorn Sangritti.
The shares were offered at 1.35 baht apiece, with subscriptions and payments conducted from Nov. 9 to Nov. 13.
The placement was fully subscribed, with no remaining shares, King Gen said.



