Z.ai released GLM-5.2, an open-source AI model featuring 750 billion parameters and a 1-million-token context window. According to NS3.AI, Z.ai shares surged more than 30% in Hong Kong trading following the announcement. The report also said the stock is up over 800% since debuting in January.

Fidelity Digital Assets has rejected concerns that Bitcoin’s long-term security will weaken as mining rewards shrink, saying the network’s incentive structure can remain strong even as block subsidies decline. According to Cointelegraph, a new research report authored by Fidelity research analyst Daniel Gray argues that Bitcoin’s security is supported by more than block rewards, pointing to transaction fees, market incentives, and other economic forces that encourage miners to keep securing the blockchain and make sustained attacks prohibitively expensive. The report challenges a longstanding criticism that each quadrennial halving undermines Bitcoin’s security by reducing new coin issuance, potentially eroding miners’ incentives unless transaction fees rise enough to compensate. The debate remains a central long-term question for Bitcoin (BTC), which follows a fixed supply schedule that gradually reduces issuance until block subsidies eventually disappear. Since April 20, 2024, miners have received a subsidy of 3.125 BTC per block, down from 6.25 BTC in the prior halving cycle. Gray said lower issuance has not translated into weaker incentives because Bitcoin’s rising price has more than offset the decline in block rewards. He cited average daily miner revenue growth from roughly $26,300 during Bitcoin’s first halving cycle to more than $40.2 million today, writing that miner incentives—and by extension network security—have historically strengthened alongside Bitcoin’s price. While Fidelity’s report maintains that Bitcoin’s long-term incentive structure remains intact, publicly traded mining companies are facing near-term financial pressure. Some industry analysts have described the current environment as among the most challenging on record, citing lower mining rewards, rising costs, and intensifying competition. In response, several miners have diversified into artificial intelligence and high-performance computing, using existing power infrastructure and data center assets to serve demand for AI workloads rather than relying solely on Bitcoin mining. A recent VanEck report estimated that publicly traded miners could require up to $50 billion in additional capital to fully transition to AI infrastructure, highlighting the scale of the shift. Blocksbridge Consulting, writing in a Miner Weekly publication, said Bitcoin mines can operate with relatively simple buildings, modular infrastructure, and ASIC fleets that tolerate fast curtailment, while AI and HPC facilities require higher standards for uptime, cooling, electrical redundancy, networking, and customer support.

Kalshi and ADI Predictstreet announced a strategic partnership to expand prediction markets globally, including co-branded placements during the FIFA World Cup 2026™ knockout stage and post-tournament international expansion. According to BeInCrypto, the deal links Kalshi with ADI Predictstreet, FIFA’s official prediction market partner, and includes a co-branded hub with football prediction markets, live updates, and fan content. The firms also plan infrastructure integrations on ADI Chain’s technology, including stablecoins, Web3 products, and on-chain settlement solutions.

Scotland coach Steve Clarke has stepped down after the team’s World Cup exit was confirmed on Saturday, with Scotland unable to progress to the last 32 as a best-placed third-place team. Clarke, 62, had signed a new deal until 2030 before the tournament in North America and leaves after leading Scotland to three major tournaments, including Euro 2021 and Euro 2024, according to Yahoo Sports.

Wall Street remained focused this week on developments in the Middle East, and energy prices eased after a ceasefire agreement was reached. According to PANews, oil prices have fallen to around $70 a barrel from $100 about a month ago. Major U.S. stock indexes are set to close out a solid first half of the year next week. The S&P 500 is up more than 7% so far in 2026, though equities faced a more difficult period in June. Gold prices saw sharp volatility this week. Early dip-buying and safe-haven demand shifted into another round of heavy selling amid stronger-than-expected U.S. economic data, persistent inflation, a stronger U.S. dollar, and rising expectations for Federal Reserve rate hikes, pushing gold back near the $4,000 level. In a late surge, gold rebounded to the edge of $4,100.

England coach Thomas Tuchel said he is concerned his right-back options are worsening after Jarell Quansah went off with a painful ankle twist in the 2-0 win over Panama, leaving Quansah and Reece James “very doubtful” for England’s last-32 tie in Atlanta on 1 July. Tuchel said Quansah has had the injury before and expects it to be “a matter of days,” while noting Tino Livramento has already been ruled out and Djed Spence replaced Quansah against Panama, according to Yahoo Sports. Jude Bellingham and Harry Kane scored seven minutes apart, and Kane’s goal was his 11th in World Cups to break Gary Lineker’s England record.

Iran’s Islamic Revolutionary Guard Corps (IRGC) said any act that violates the ceasefire agreement would also violate the Islamabad agreement and would lead to a complete halt of all ongoing processes. According to Odaily, Iran’s Press TV reported that the IRGC said it may take tougher measures against vessels in the Strait of Hormuz.

金色周刊是金色财经推出的一档每周区块链行业总结栏目，内容涵盖一周重点新闻、行情与合约数据、矿业信息、项目动态、技术进展等行业动态。

巨头下场做AI员工，创业公司先被验证还是先被挤压。

BTC 承压行业收缩，监管政策催化传统金融加速整合加密赛道。

将这半年看似分散的动作连起来看，中国央行在修的，其实不只一条跨境支付的路，而是一整套能让数字人民币真正运行起来的环境。

人工智能,技术,数据飞轮还是重复样本？Physical AI 该告别「小时数崇拜」了 金色财经,机器人公司还在拼小时数，但真正缺的是新样本。

早在 2016 年，SBI 就和 Ripple 达成了合作。

隔夜美股，全球半导体板块遭遇猛烈抛售。费城半导体指数大幅走低，美光科技和闪迪盘中跌幅均超6%，西部数据跌逾7%。

据媒体报道，美国政府周五宣布解除对该公司Claude Mythos 5模型的出口限制，允许超过100家获政府批准的企业和联邦机构访问该模型。

这套机制究竟是“稳定器”还是“加速器”，答案就藏在接下来的价格与利率里。

人工智能行业持续高歌猛进，丝毫没有降温的迹象。然而，“区块链 AI”领域的情况却截然不同。为什么它未能吸引到相当的关注？